A non-dispersion strategy for large-scale production of ultra-high concentration graphene slurries in water

It is difficult to achieve high efficiency production of hydrophobic graphene by liquid phase exfoliation due to its poor dispersibility and the tendency of graphene sheets to undergo π−π stacking. Here, we report a water-phase, non-dispersion exfoliation method to produce highly crystalline graphene flakes, which can be stored in the form of a concentrated slurry (50 mg mL−1) or filter cake for months without the risk of re-stacking. The as-exfoliated graphene slurry can be directly used for 3D printing, as well as fabricating conductive graphene aerogels and graphene−polymer composites, thus avoiding the use of copious quantities of organic solvents and lowering the manufacturing cost. This non-dispersion strategy paves the way for the cost-effective and environmentally friendly production of graphene-based materials.

(this is logical in light of the oxidative treatment). Thus, the Raman spectrum represents the multi-layered graphene flake, where the top layer is fully oxidized, i.e. the top layer is graphene oxide. It is very much the same GNP as the product obtained in ref. [25], authors citing.
The question: if the authors obtain mostly single-layered graphene, as they show on the AFM images (Fig. 2,e;   Fig. S2; Fig. S3), then why they picked the multi-layered oxidized flake for the Raman mapping? All the data should be in accordance with each other.
3. Based on Fig. 2b, I would expect the flakes to be thicker along the perimeter. The AFM height profiles for GO normally show 0.9-1.2 nm height. Graphene -0.5-0.6 nm. Could the authors, please, comment on this in the manuscript text? If the flakes are indeed thicker on the edges, as I can see from the provided height-profiles for some flakes, discussing this in the text will strengthen the publication. I would recommend to acquire a high resolution image of one single flake and carefully examine it.
4. The SEM images of as-made graphene ( Fig. 4b and Fig. S1c) also look more like those for multi-layered GNP.
Thus, the two data (Raman and SEM) provided by the authors, suggest that they obtain GNP, but not singlelayered graphene, as they claim. The 3D priniting and the fabricated aerogels do not confirm the single-layer character.
The complete exfoliation of graphite source to the signle-layer graphene is the main claim of this study that makes it, according to the authors, advantageous over the previous works. Thus, this needs to be unambiguously confirmed.
I will be happy to support after addressing these questions. Major revision is suggested.
Line 220: "To the best of our knowledge, this is the first demonstration of water-phase 3D-printing of exfoliated graphene 43,44 . Previous efforts on graphene aerogels rely on the sol-gel chemistry, which is challenging for large-scale production 45,46 ." Line 221: "The electrical conductivity reaches ~ 197 S m -1 at a density of 100 mg cm -3 (Fig. 5g), which is comparable to 3D-printed rGO networks by conventional dispersion approaches 42,43 although it is inferior to that of CVD-grown method 44 . By incorporating high-temperature annealing in commercial graphite production, it is possible to further improve the electrical conductivity to a level comparable with that of CVD-grown aerogels 48 ." Line 224: "The printed graphene aerogels can be used as 3D templates for in-situ polymerization, which are promising for applications in a wide range of energy storage devices 49 and durable absorbent materials 50 ."

Reviewer #2:
The manuscript claims the processing graphene using an aqueous system for processing well dispersed graphene sheets up to 23 wt%. The processing method utilized high speed shearing of pre-treated graphite powders to exfoliate the sheets. Jamming interactions between graphene sheets prevents the sheets from aggregating during exfoliation at pH = 14. The pre-treatment of graphite involves exposing graphite to KMnO4 and concentrated H 2 SO 4 , which partially oxidizes graphite layers. While partial oxidation generate enough ionic repulsion to improve exfoliation and stability, the π-π conjugated graphene structure remains intact for electronic conduction. As a demonstration for the slurry's application, the slurry was applied as a 3D printable gel, which can be further freeze-dried into porous scaffold. Overall, this is an excellent piece of work.
Question 1: The article demonstrated a significant improvement to solution exfoliation of graphene. By partially oxidizing graphite particles, liquid exfoliation of graphene can be scaled up without significantly sacrificing graphene quality. This is a much sought after improvement to the preparation of graphene.

Response:
We thank the reviewer for his/her positive comments on our paper. We have addressed his/her concerns on the partially oxidized graphite, control experiments on GO solution and the recovery of sulfuric acids in this revision.
Question 2: Line 60, the authors refer to pre-treated graphite as "sulfate-intercalated graphite". However, the graphene sheets' pH response and elemental analysis indicate that the sheets were partially oxidized. The pre-treatment method is also similar to graphene oxide synthesis with variation in molar ratios. Wouldn't it be better to call it "partially oxidized graphite" Response: We agree that "sulfate-intercalated graphite" is confusing since we employed a small amount of oxidizing agent (KMnO 4 ) during intercalation. It is known that oxidizing agents, even when used in a low molar ratio (1 wt. equiv. in this manuscript) and low temperature, can result in a certain degree of oxidation on graphite flakes (5.9 atom% O) (The chemistry of graphene oxide, Chem. Soc. Rev., 2010, 39, 228-240; Chemistry with graphene and graphene oxide -challenges for synthetic chemists, Angew. Chem. Int. Edit., 2014, 53, 7720-7738). In this regards, we have modified the "sulfate-intercalated graphite" into "partially oxidized graphite" throughout this manuscript.
Line 60: "Sulfate-intercalated Partially oxidized graphite is used as the precursor, which is exfoliated by high-rate shear in an alkaline aqueous solution of pH = 14." Line 77: "Pristine graphite was partially oxidized using a very low molar ratio of oxidizer to carbon in graphite (0.076) in order to generate a low density of ionizable oxygen-containing groups on graphene layers." Line 107: "…confirming that the crystal structure of graphene is well retained after sulfate-intercalation partial oxidation and shear-exfoliation." Line 248: "Pretreated graphite was obtained using the conventional intercalation process using sulfuric acid." Question 3: Similarly for lines 79-81, partial oxidation would yield a similar shift in XRD spectrum.

Response:
We agree that partial oxidation would yield a similar shift in XRD spectrum. Partial oxidation and graphite intercalation may proceed concurrently during the oxidation-intercalation of graphite. We have modified the term of "sulfate-intercalated graphite" into more accurate "partially oxidized graphite" throughout this manuscript.
Line 79: "Partial oxidation induced peak at 22.5 o in the X-ray diffraction (XRD) spectrum indicates the formation of a stage-1 graphite intercalation compound with an interlayer distance of 8.0 Å ( Supplementary  Fig. 6) 27,28 ." Question 4: The reported graphene slurry has a high concentration of flocculating single-layer graphene sheets at pH of 14, and authors compared this mixture to GO solutions with well dispersed nano-sheets. A more direct comparison is between GO solution and graphene slurry at pH = 12, the stable pH for storing and utilizing graphene slurry.

Response:
We have supplied new control experiments on the GO solution and graphene slurry at pH = 12 in the revised version. Both GO and graphene dispersion at pH = 12 can be stored for weeks without obvious aggregations in Supplementary Figure 7 (a, b). The maximum dispersion concentration is similar to those reported in the literature (~ 0.1 mg mL -1 for graphene in water), whereas our concentrated graphene slurry (pH = 14) can be re-dispersed on-demand. To further examine its stability for practical applications, we have also examined the TEM image and Zeta-potential of graphene dispersion at pH = 12 after storing for 1 week. As shown in Supplementary Figure 7 (c), the graphene flakes are loosely packed into few-layer morphology after solvent evaporation. We can observe many NaOH absorbates on graphene flakes in TEM image, which contributes to the surface potential against π-π re-stacking. Additionally, the Zeta potential of graphene dispersion after 1 week (-42.4 mV) is very close to the value of fresh solution (-48.4 mV), indicating excellent stability of graphene dispersion in this condition. Related discussions have been included in the Supplementary Information. Figure 7. (a,b) Digital photos of graphene and GO dispersion at pH = 12 after storing for a week, both at 0.1 mg mL -1 . (c) TEM image of graphene dispersion at pH = 12 after a week.

Question 5: At line 253, the authors claim that sulfuric acid can be recovered and recycled. This may be easier said than done. Can the authors demonstrate this or provide relevant references to support this?
Response: To address the reviewer's concern, we have supplied a video recording the recovery of sulfuric acid in the revised version (Supplementary Video 1). Details on the experimental setups can be found in Supplementary

Reviewer #3:
The manuscript by Dong et al. proposes the new approach for the liquid exfoliation of graphite to yield graphene. To achieve better exfoliation, the authors subject graphite to partial oxidation with potassium permanganate is sulfuric acid, and then subject the as-obtained product to shear in aqueous solution of NaOH at pH=14. The difference from the previous studies is that the product is obtained not as stable dispersion, but in the form of flocculated aqueous slurry with concentrations as high as 5 wt%. The presence of adsorbed ions prevents the irreversible restacking of graphene flakes and enables their re-dispersion in solution on demand. The authors further demonstrate the use of the as-made graphene slurry for 3D printing. This approach is novel, and has great potential for practical applications, since avoiding copious amounts of water and/or organic solvents helps one to lower the manufacturing cost, and to minimize expenses with further storage and processing.
The only concern I have relates to the experimental data, confirming the quality of as-obtained graphene.

Response:
We thank the reviewer for his/her positive comments on our paper. We have addressed his/her concerns on the quality of as-obtained graphene, especially on the Raman spectra of graphene sheet in the revision. As shown in the latter study, at this graphite/oxidizer ratio, roughly one half of the body of a graphite flake is converted to graphite oxide. The question: what authors do differently, that they obtain almost nonoxidized graphene? Shorter exposure time, lower temperature, etc.? Explanation for this apparent discrepancy with the literature data should be given.

Response:
We appreciate the critical comments on graphite oxidation. In comparison with the listed (Ref [37][38][39][40][41] and related references, there're several reasons for the much lower oxidation degree in our case: 1) we use graphite with bigger size (100 mesh ≈ 150 μm), which can significantly reduce the diffusion and intercalation of sulfuric acid and oxidant into the interlayer spacing of graphite; 2) we use a much lower oxidant to graphite ratio (1 wt. equiv. of KMnO 4 for initial feeding and ~0.8 wt equiv. of KMnO 4 for finial consumption) to avoid over-oxidation, while GO preparation usually requires over 3 wt. equiv. of KMnO 4 together with 0.5 ~ 1 wt. equiv. of NaNO 3 (for in-situ generation of HNO 3 ); 3) the hydrolysis of graphite intercalated compounds has been identified as a key step in graphite oxidation (J. Am. Chem. Soc. 2012, 134, 2815). Conventional GO preparation involves water injection into the reaction mixture (conc. H 2 SO 4 etc.), this causes a sudden temperature increase which facilitate the oxidation of graphite.
In contrast, we recover the excess sulfuric acids prior to hydrolysis and pour the intercalated compounds into a large amount of ice-water (20 wt. equiv.) to minimize the heat generation from concentrated sulfuric acids during hydrolysis. We do not keep the hydrolysed compounds for further reaction. Excess amount of 30 wt% H 2 O 2 is injected to terminate the reaction immediately; 4) the reaction temperature is low (r.t.) with a short reaction time (2 h).
A detailed comparison is presented in Table 1 and Supplementary Table 3 It is also necessary to explain why we choose such experimental parameters for partial oxidation. As shown in Supplementary Figure 13, we examined the influence of KMnO 4 wt. equiv. on the final yield and quality of graphene. We found that 1 wt. equiv. of KMnO 4 are optimized for both yield and quality, whereas using a lower amount of KMnO 4 results in insufficient intercalation and a much worse graphene yield. Raman analysis reveals that for treatment conditions using low oxidant to graphite ratios (0.4 or 1.0 wt. equiv.), the aromatic structure of graphene is well-reserved with a low I D /I G ratio. Further increase in oxidant usage will dramatically reduce the quality of exfoliated graphene and finally convert graphene into graphene oxide. Based on the above observations, the optimized conditions (1 wt. equiv. KMnO 4 etc) were used in this manuscript to reduce unnecessary oxidations. Figure 13. (a) Graphene yields are greatly influenced by the ratio of oxidant to graphite, suggesting that appropriate oxidation is essential; (b) Raman spectra of graphite at various KMnO 4 ratios, together with a comparison of GO. Inset of (a), photo of graphite at different KMnO 4 ratio, (I) 0.4 wt. equiv., (II) 1 wt. equiv., (III) 2 wt. equiv. and (IV) 4 wt. equiv. Question 2: The Raman mapping (Fig. 2b) does not match the rest of the provided experimental data. The AFM and TEM images show single-layer graphene. The Raman spectrum does not represent the single-layer graphene flake. This is apparent from the G/2D ratio and the character of the D and G-bands. The Raman spectrum of a single-layered graphene is very different: 2D signal must be higher than G-band. Here the situation is opposite, strongly attributing the given spectrum to multi-layered graphene (or graphite) with number of layers >5. Both D-band and G-band are broadened, suggesting highly damaged structure, with the density of defects similar to that in GO (this is logical in light of the oxidative treatment). Thus, the Raman spectrum represents the multi-layered graphene flake, where the top layer is fully oxidized, i.e. the top layer is graphene oxide. It is very much the same GNP as the product obtained in ref. [25], authors citing. The question: if the authors obtain mostly single-layered graphene, as they show on the AFM images (Fig. 2,e; Fig. S2; Fig. S3), then why they picked the multi-layered oxidized flake for the Raman mapping? All the data should be in accordance with each other.

Response:
In order to collect AFM and TEM data, we spin/drop-coated single-layer sheets aqueous dispersion on mica substrate/TEM grid and hence the data is representative of monolayer flake. Our Raman was collected from the dried, aggregated samples. We have redispersed the samples in an attempt to collect better quality Raman that is representative of monolayers, however, the best data in Figure 2 show an I 2D /I G ratio of ~ 0.45, corresponding to ~ three-layer graphene (Raman spectroscopy in graphene, Phys. Rep., 2009, 473, 51-87; Raman spectrum of graphene and graphene layers, Phys. Rev. Lett., 2006, 97, 187401). The data we have obtained still lead us to conclude that what we have is predominantly single layer graphene because of the following reasons: 1) We used Si wafer for Raman measurements since the transparent mica shows low optical contrast and cannot be used for sample tracing. Due to the different wettability of Si wafer and mica, we have to use NMP to re-disperse and spin-coat graphene onto Si wafer. This leads to a huge problem in getting isolated graphene sheets because of the high-boiling point of NMP and serious coffee-ring effect (Bi-and trilayer graphene solutions, Nat. Nanotech. 2011, 6, 439-444). Most graphene sheets are stacked with each other.
2) It is well-known that defects have a strong influence on the Raman spectrum of graphene.
Although the 2D band is not so sensitive to the defective sites as the D band, it is possible that the intensity of the 2D band becomes reduced after oxidation, functionalization or other treatments (Spectroscopy of covalently functionalized graphene, Nano Lett. 2010, 10, 4061-4066). In fact, for solution exfoliated graphenes, it is generally difficult to obtain a similar I 2D /I G ratio as defect-free, mechanically exfoliated graphenes. A detailed comparison between our Raman data and the literature is provided below, from which we can find our Raman data provide a relatively high I 2D /I G ratio for the solution exfoliated graphenes. In this regard, although we are not able to provide the Raman spectrum mentioned by the Reviewer, we believe it is a technical issue related to the processing of Raman measurements.
To clarify this mismatch between Raman and AFM/TEM, we have revised the Main Text Figure 2b and added a short comments in the revision:   Response: According to the reviewer's suggestion, we have provided a high resolution AFM image in Figure 4 to examine the thickness at the edge and the basal plane. Basically, we do not observe such a different thickness in our sample, which can be explained by the following reasons: 1) We used tapping AFM to acquire the images and fluctuation in edge thickness is due to instrumental error and can be corrected by slowing down the scan speed of the AFM. The AFM line scan in Main Text Figure 2d is collected relatively fast and over a large scan area (~ 25 μm), resulting in a large instrument error (lower thickness at basal plane). However, when we performed a high resolution AFM scan, such fluctuation at the edge is quite small for our graphene (Supplementary Figure 2), indicating that the thickness at the edge is almost similar to that of the basal plane; 2) It is a well-known phenomenon that water (or small molecules) is trapped between graphene and substrate, resulting in an apparent higher thickness (e.g., 0.9 -1.2 nm for graphene oxide) compared to the monolayer thickness of graphene (0.34 nm). Partial intercalation at the edge is difficult due to the large local strain. Thus, we observed similar thickness (0.5 ~ 0.6 nm) at the edge and basal plane, we do not think we need to discuss this at length.

Question 4:
The SEM images of as-made graphene ( Fig. 4b and Fig. S1c) also look more like those for multi-layered GNP. Thus, the two data (Raman and SEM) provided by the authors, suggest that they obtain GNP, but not single-layered graphene, as they claim. The 3D printing and the fabricated aerogels do not confirm the single-layer character.
The complete exfoliation of graphite source to the single-layer graphene is the main claim of this study that makes it, according to the authors, advantageous over the previous works. Thus, this needs to be unambiguously confirmed.

Response:
We thank the reviewer for the critical comments. Main Text Figure 4b and Supplementary Figure 1 (c) in the original manuscript were obtained from freeze-dried graphene powders, where aggregation was unavoidable. We have supplied new SEM image by spin-coating graphene NMP dispersion onto Si substrate to minimize such aggregation. We also compare our SEM images with the literature (Figure 5), where our graphene nanosheet has a similar morphology to the "single-layer" graphene. Owing to the coffee-ring effect during solvent evaporation (Nat. Nanotechnol. 2011, 6, 439; J. Phys. Chem. C 2014, 118, 27081), it is a natural phenomenon that re-stacking of graphene nanosheets occurs. Thus, it is challenging to determine the single-layer morphology solely from the SEM images.
We employed two additional techniques: AFM and TEM/SAED, to validate that our graphene nanosheets are single-layer. As shown in Figure 2 (d) in manuscript, the height profile together with a statistical analysis of over 100 flakes show that > 90 % of the flakes are single layer (<1 nm in thickness) in AFM. The characteristic SAED pattern and new highresolution TEM images in Supplementary Figure 3 and 4 also confirm the single-layer morphology, although we can occasionally find some two or three-layer graphene (Supplementary Figure 4 (h), (i)). To explain the mismatch between Raman and AFM, we also supply new Raman spectra in the revised version. Please refer to our response to your Questions 2, 3.