Activated Carbons from Hydrochars Prepared in Milk

Hydrothermal carbonization converts organics in aqueous suspension to a mixture of liquid components and carbon-rich solids (hydrochars), which in turn can be processed into activated carbons. We investigated whether milk could be used as a medium for hydrothermal carbonization, and found that hydrochars prepared from milk, with or without an added fibrous biomass, contained more carbon (particularly aliphatic carbon), less oxygen, and more mineral components than those prepared from fibrous biomass in water. Activated carbons produced from hydrochars generated in milk had lower specific surface areas and CO2 capacities than those from hydrochars formed in water; however, these differences disappeared upon normalizing to the combustible mass of the solid. Thus, in the context of N2 and CO2 uptake on activated carbons, the primary effect of using milk rather than water to form the hydrochar precursor was to contribute inorganic mass that adsorbed little CO2. Nevertheless, some of the activated carbons generated from hydrochars formed in milk had specific CO2 uptake capacities in the normal range for activated carbons prepared by activation in CO2 (here, up to 1.6 mmol g−1 CO2 at 15 kPa and 0 °C). Thus, hydrothermal carbonization could be used to convert waste milk to hydrochars and activated carbons.

Milk is produced on an enormous scale, and as a result, so is waste milk. In Europe, 13% of milk produced is wasted, and in North Africa and West and Central Asia, the value is 20% 1 . In both regions, 3.5% of the milk produced is wasted at the production phase 1 , where it could potentially be recovered relatively easily. Even in Sweden, where an exceptionally low percentage of milk is wasted at production, the amounts of milk waste generated are large. For example, 0.32% of milk produced at Swedish farms in 2011 was discarded at the farm, primarily following antibiotic treatment of the cows for mastitis; this amounts to more than 9000 tons of milk 2 . Related to the issue of waste milk is dairy wastewater, which is composed of milk as well as additional water and detergents used for cleaning and sanitizing equipment 3 .
Milk is an aqueous dispersion (in the case of homogenized milk) or suspension (non-homogenized milk) of fats, proteins, and sugars, and also contains inorganic cations including K + , Na + , Ca 2+ , and Mg 2+4 . Aqueous preparations of organics, including suspensions of biomass, can be converted to carbon-rich solids called hydrochars via hydrothermal carbonization, i.e. by heating to (typically) 180−250 °C under autogeneous pressure [5][6][7][8][9][10] . The ability to convert wet biomass is the main process-related advantage of hydrothermal over pyrolytic carbonization 7 . Yoghurt (10 wt% in water) has been converted to a hydrochar that was evaluated as fuel 11 . Additionally, the hydrothermal carbonization of milk has been used as the first step in the synthesis of antibacterial carbon dot−Ag nanoparticle composites 12 .
To evaluate hydrothermal carbonization as a method of using waste milk, we converted homogenized milk to hydrochars that were characterized and activated to give activated carbons ( Fig. 1) whose properties and CO 2 sorption abilities were measured. Milk was also studied as a medium for the hydrothermal carbonization of fibrous biomass; thus, corn husk or flax fiber (corn husk is a waste product, and flax fibers are relevant to Swedish

Results and Discussion
Hydrochars. The hydrothermal carbonization of flax fiber and corn husk in water at 220 °C for 24 h gave hydrochars in 40 and 34% yield, in line with yields obtained in other studies of hydrothermal carbonization at moderate temperature and extended times 6 . More hydrochar was obtained when homogenized milk was used as the liquid for hydrothermal carbonization. The solid yields from hydrothermal carbonization in milk can be estimated by taking the combined mass of added solid plus solid components in the milk as the solid input; using this method, the yields from the hydrothermal carbonization of flax fiber and corn husk in milk were 81% and 70%.
The hydrochars produced in milk contained more H and N, but less O, than their counterparts produced in water (Table 1). The greater H content was reflected in the IR spectra of the HC-xx-M ( Supplementary  Fig. S1a), which showed much more intense ν(C−H) bands, primarily associated with aliphatic C−H bonds (3000−2800 cm −1 ), than the spectra of the HC-xx-W. The 13 C NMR spectrum of HC-CH-W (Fig. 2a) resembled that reported for HC produced from rye straw at 240 °C 27 , showing peaks for both saturated (δ < 80 ppm) and unsaturated (δ > 100 ppm) carbons, as well as unsaturated oxygenated groups such as carboxylic acids (δ ~ 175 ppm) and ketones (δ ~ 205 ppm). A very small peak at δ ~ 72 ppm may have indicated the presence of unreacted sugars or cellulose 27,28 . Extraction of HC-CH-W in acetone lowered the 13 C NMR intensity associated with saturated carbons, in particular for the peak at δ ~ 30 ppm, and the concomitant loss of carboxylic acid and ketone carbons suggested that levulinic acid was a component of the extractable material, which was obtained as a darkly colored solid. Free levulinic acid has been detected in hydrochars from glucose 29 . In agreement with the IR results, the 13 C NMR spectrum of HC-CH-M ( Fig. 2b) revealed it to contain a much larger fraction of saturated carbons than HC-CH-W. Further, the fraction of the carbons that were oxygenated (δ ~ 150 and ~ 50 ppm for unsaturated and saturated carbons) was lower in the hydrochar produced in milk, consistent with the lower n O /n C ratio observed for the hydrochars generated in milk (Table 1) 29 . After extraction with acetone, the fraction of saturated carbons in HC-CH-M fell, and the extract itself was a dark viscous oil. Fatty acids are not readily converted to hydrochar, but do adsorb onto hydrochars formed from sugars 30 ; they can then be extracted using ethers 30,31 or ethanol 32 . Therefore, the extractable saturated carbons on HC-CH-M were likely largely from fatty acids. The HC-xx-M had lower surface areas than the HC-xx-W (Table 1), and this difference is attributed to extractable molecules adsorbed on and in the pores of the HC-xx-M.
The HC-xx-W were more thermally stable at lower temperature, losing less than 5% of their mass when heated in air over 100-250 °C, whereas the HC-xx-M lost 15-20% of their mass in the same temperature range ( Supplementary Fig. S1b). Heating HC-xx-W to 800 °C in air left very little residue (<0.6 wt%), whereas the hydrochars produced in milk retained 3-8% of their mass ( Table 1), indicating that some of the mineral elements from the milk were retained. The IR spectra ( Supplementary Fig. S1a) of the hydrochars produced in milk showed two peaks, at approximately 600 and 560 cm −1 , that were not observed for the hydrochars produced in water. The positions and relative intensity of these peaks are consistent with those for vibrations associated with the phosphate groups of apatite 33 , and they may therefore be related to an inorganic phosphate.
Activated carbons. The as-synthesized hydrochars were heated at 800 °C in CO 2 for 4-20 h to give activated carbons. Very high capacities for CO 2 sorption have been observed for hydrochars after activation with KOH 15 or K 2 CO 3 13 ; however, the use of solid etchants requires an additional washing step in the material preparation, and KOH in particular is corrosive 34 , and we therefore focused on activation with CO 2 . Generally, hydrochars were heated in CO 2 and then held at 800 °C under CO 2 , but a modified procedure was also tested for a few samples. Here, the solid was heated to 800 °C under N 2 before the gas stream was changed to CO 2 and the sample held at 800 °C for 20 h. The resulting activated carbons are distinguished with the term 'N 2 CO 2 ' in the sample name.  www.nature.com/scientificreports www.nature.com/scientificreports/ Scanning electron microscope images of activated carbons derived from the HC-M sample produced with no solid biomass (Fig. 3) showed two types of particles. The smaller particles (~5−50 μm; Fig. 3a,c,e) seemed smooth and were agglomerations of spheres that were reminiscent of the carbonaceous spheres seen in the hydrothermal carbonization of carbohydrates 27,[35][36][37][38] , along with more irregular macroporous particles. Energy dispersive X-ray spectroscopy (EDS) of one such particle ( Supplementary Fig. S2a) in AC-M-4 indicated that it was composed primarily of carbon and oxygen, but also contained small amounts of calcium, phosphorus, potassium, and magnesium. Larger (hundreds of μm), irregularly shaped particles with defined edges, sometimes bearing small spheres on their surfaces, were also present (Fig. 3b,d,f), and EDS showed one such particle to be composed of iron and www.nature.com/scientificreports www.nature.com/scientificreports/ oxygen, and to a lesser extent carbon ( Supplementary Fig. S2b). We have previously observed Fe in activated carbons generated from hydrochars, even when no Fe precursor was added; this is derived from the stainless steel reactor used during the activation 39 .
The surface morphologies of flax fiber and corn husk ( Supplementary Fig. S3a,b) were retained throughout hydrothermal carbonization and activation, with AC-FF-W-10 ( Supplementary Fig. S3c) appearing as short fibers, and AC-CH-W-10 ( Supplementary Fig. S3d) as broader sheets. These structures were also retained when milk was used in the hydrothermal carbonization ( Supplementary Fig. S3e,f), but in that case were accompanied by the carbonaceous spheres and amorphous material seen in the activated carbons produced without solid biomass (Fig. 3).
Activation increased the aromaticity of the carbonaceous solids, as n H /n C fell in all cases (from 0.79-1.3 for hydrochars to 0.15-0.34 for activated carbons; cf. Tables 1 and 2). The formation of partially graphitized carbon was evinced by broad X-ray diffraction (XRD; Fig. 4 and Supplementary Fig. S4) peaks centered at 2θ = 23-25 and 43°, which correspond to the (002) and (10) planes of turbostratic carbon 40 , for most samples. These peaks were very weak for samples derived from milk without an additional biomass source (i.e. AC-M-t, Supplementary Fig. S4e), and for samples heated to 800 °C in N 2 (Fig. 4, cf. Supplementary Fig. S4). In two samples that were examined with X-ray photoelectron spectroscopy (XPS of AC-CH-y-N 2 CO 2 -20 for y = W and M; see Supplementary Figs S5 and S6), the C 1s peaks included long slopes toward high binding energies, which supported the presence of graphitic or carbon black-type structures, though detailed deconvolution was not possible.
Whereas activation increased the carbon content of hydrochars formed in water (from 66-71 wt% C for HC-xx-W to 73-94 wt% C for AC-xx-W-t; cf. Tables 1 and 2), it decreased carbon content for most of the HC formed in milk (from 62-73 wt% C for HC-xx-M and HC-M to 35-75 wt% C for AC-xx-M-t and AC-M-t), because the removal (gasification) of organic material during activation caused the mineral components from the milk to make up a larger fraction of the activated carbons. Thus, although the residual mass of the AC-xx-W-t samples after combustion to 800 °C was never greater than 12%, it ranged from 11-48% for the AC-xx-M-t samples, with high values being observed particularly for samples with long activation times ( Table 2).
There are likely multiple reasons for the larger mineral content of the activated carbons produced from milk-derived hydrochars as compared with activated carbons prepared from hydrochars generated in water. First, there were minerals in the milk, and hence more mineral components were observed in the hydrochars prepared from milk than from water, as expected (see above). Further, the yield from activation of an HC-xx-W was generally higher than that from activation of HC-xx-M when the solid feedstock and activation conditions were the same ( Table 2); that is, more mass was removed from HC-xx-M. Thus, either the HC-xx-M were more readily etched than the HC-xx-W, or the mineral components from milk catalyzed the decomposition of the hydrochars, or both. We cannot reject the former hypothesis, as HC-xx-W and HC-xx-M were chemically different (  Table 2. Properties of the activated carbons prepared from the activation of hydrochars generated in water or milk. a Non-combustible mass = percentage of mass that remains after heating the sample to 800 °C in 25 mL min −1 air. b S BET = Brunauer-Emmett-Teller surface area 54 , calculated from the N 2 adsorption isotherms ( Supplementary Fig. S7) over P/P 0 = 0.01-0.1. c V μ-pore = micropore volume, calculated from the N 2 adsorption isotherms ( Supplementary Fig. S7) using the Dubinin-Radushkevich equation 55,56 fitted over P/P 0 = 0.0001-0.05. d CO 2 uptake at 101 kPa is measured (isotherms in Supplementary Figs S9-S13); CO 2 uptake at 15 kPa is interpolated from a two-site Langmuir fit to the isotherm data (details in Supplementary Information section S1.4). e No solid biomass source was added.
www.nature.com/scientificreports www.nature.com/scientificreports/ Fig. 2 and Supplementary Fig. S1); however, the metal ions present in milk likely also affected the process. Ca 2+ catalyzes the gasification of biochars in CO 2 41 , and K + salts including KOH 15,35,42,43 , K 2 CO 3 13,44 , KHCO 3 45,46 , and K 2 C 2 O 4 47 are used to activate hydrochars and form activated carbons. Na + salts can also be used in the preparation of activated carbons from hydrochars 9 . Elemental analysis of AC-FF-W-10 revealed no detectable K and only 0.03 wt% Ca; whereas AC-CH-M-10 contained 0.34 wt% K and 2.7 wt% Ca. Similarly, XPS (Supplementary Table S1) showed that AC-CH-M-N 2 CO 2 -20 contained K and Ca, whereas AC-CH-W-N 2 CO 2 -20 did not. Thus, the Ca 2+ and K + in milk were at least partially retained throughout hydrothermal carbonization and activation, and likely contributed to pore development in the activated carbons produced from HC-xx-M and HC-M.
The primary crystalline calcium-containing phases ( Fig. 4 and Supplementary Fig. S4) in the activated carbons prepared from HC-xx-M were calcium phosphate (ICSD 00-003-0713) or calcium-rich mixed calcium magnesium phosphates, such as Ca 19 Table S1) showed no Mg, so the more Mg-rich Ca 19 Mg 2 (PO 4 ) 14 is less likely to be important. Some potassium may have been present as K n Na 1−n Cl (n = 0.2, ICSD 01-076-3440; n = 0.0997, ICSD 01-075-0305), but the small amounts involved and the presence of other phases render this assignment uncertain. Neither the calcium (magnesium) phosphates nor K n Na 1−n Cl were significant in the activated carbons prepared from HC-xx-W. Two significant non-carbon phases were observed for activated carbons derived from hydrochars formed in both water and in milk; these were α-Fe (ISCD 01-071-4648), which gave rise to sharp peaks at 2θ = 44 and 65°, and an iron oxide (cubic Fe 3 O 4 , inverse spinel Fe 3 O 4 , or γ-Fe 2 O 3 ; these produce similar powder XRD patterns), consistent with the observation of an iron oxide by EDS (Supplementary Fig. S2). Nevertheless, based on XPS the total Fe in the activated carbons was small (Supplementary Table S1).
A consequence of the high inorganic content of the AC-xx-M-t was that they displayed lower apparent specific surface areas S BET (300-480 m 2 /g) than the AC-xx-W-t (400-750 m 2 /g; Table 2, Fig. 5a; N 2 sorption isotherms Supplementary Fig. S7). S BET was correlated to activation time, but even the milk-derived activated carbons with the longest activation times had lower S BET than most of the AC-xx-W-t. This difference did not reflect large discrepancies in the S BET values of the carbonaceous portions of the activated carbons. Rather, when S BET values were normalized to the combustible mass (fraction of mass lost upon heating to 800 °C in air) of each AC (Fig. 5b), there was no consistent difference between the activated carbons produced from HC-xx-W and HC-xx-M, though activation time remained a significant determinant of S BET . Thus, in terms of S BET , the primary impact of using milk as a starting material was to contribute low-surface-area inorganic mass.
AC-xx-M-t had different pore structures than AC-xx-W-t ( Supplementary Figs S7 and S8). All of the activated carbons contained micropores, as indicated by N 2 uptake at low pressure, but some also contained mesopores, as revealed by hysteresis in N 2 uptake from P/P 0 ~ 0.45. Activated carbons generated from HC-xx-M, even using shorter activation times, were mesoporous, especially when no solid precursor was used in the hydrochar (i.e. for the AC-M-t samples). This difference was likely due to the catalytic effect of the mineral components in HC-xx-M in etching the carbon; a larger average pore size has been observed in polymer-derived activated carbons when Ca 2+ was added prior to activation 26 . Micropore volume (V μ-pore ) increased with activation time, and the AC-xx-W-t samples consistently had higher V μ-pore than the analogous AC-xx-M-t samples (Fig. 6a and Table 2). As was the case for the S BET , the difference in the V μ-pore between AC-xx-W-t and AC-xx-M-t disappeared upon normalizing to the combustible mass of the AC (Fig. 6b). www.nature.com/scientificreports www.nature.com/scientificreports/ As V μ-pore is an excellent predictor of the CO 2 sorption capacity of activated carbons, particularly under atmospheric CO 2 pressure 48 , we expected the activated carbons generated from HC-xx-M to take up less CO 2 than those from HC-xx-W. Indeed, the AC-xx-M-t generally took up less CO 2 than the corresponding AC-xx-W-t (Supplementary Figs S9-S13, Table 2), both at 15 and 101 kPa, and CO 2 uptake was correlated to V μ-pore , particularly at 101 kPa CO 2 ( Supplementary Fig. S14). Nevertheless, AC-CH-M-10 and AC-FF-M-10 each took up more than 1.6 mmol g −1 CO 2 at 15 kPa and 0 °C, which is typical for activated carbons generated via activation in CO 2 ( Table 3). For example, it is in the range observed for activated carbons generated from the CO 2 -activation of other waste-derived hydrochars 19 , though lower than for activated carbons generated by CO 2 -or steam-activation of isolated 49 or chemically modified 50 cellulose.  www.nature.com/scientificreports www.nature.com/scientificreports/ Overall, the best predictor of CO 2 uptake capacity in the activated carbons produced here, from hydrochars generated in water or in milk, was their carbon content (Fig. 7). The correlation between C content and CO 2 uptake was particularly strong at P CO2 = 15 kPa (Fig. 7a), the partial pressure of CO 2 relevant to flue gas cleaning. Thus, despite that V μ-pore increased with activation time (Fig. 6), the relationship between activation time and CO 2 uptake was more complex (Supplementary Fig. S15), especially for activated carbons derived from hydrochars generated in milk. The HC-xx-M lost more C atoms (and thus had greater concentrations of inorganics) upon extended 20-h activation (Table 2), so the highest CO 2 uptakes on AC-xx-M-t were obtained for activated carbons that had been activated for 10 h. Consistent with the dependence of CO 2 capacity on the carbon content of the activated carbon, the CO 2 uptake on AC-xx-W-t and AC-xx-M-t were not systematically different after normalizing to combustible mass. These values were clearly influenced by activation time though; longer times gave higher CO 2 uptake per unit combustible mass at P CO2 = 101 kPa, but the opposite was true for CO 2 uptake at P CO2 = 15 kPa (Supplementary Fig. S16). This difference can be understood in terms of pore development. CO 2 uptake at low pressure depends on the volume of very small micropores (d ≤ 0.5 nm), whereas even larger micropores (d ≤ 1 nm) are important for CO 2 uptake at P CO2 = 101 kPa 48 . Activation in CO 2 for extended times produces more volume in larger micropores and less in smaller micropores 51 , and thus benefits CO 2 uptake at P CO2 = 101 kPa.
The heats of adsorption Q st for CO 2 on the activated carbons produced from HC-xx-W and HC-xx-M (generally, Q st = 22−32 kJ mol −1 ; Supplementary Fig. S17) were consistent with the values for the physisorption of CO 2 on similar activated carbons. They were in the range observed on activated carbons derived from polymers pyrolyzed in the presence of KOH 52 as well as on polymer-derived activated carbons containing CaO nanoparticles 26 , slightly higher than the values measured on a commercial NORIT activated carbons at similar loadings 53 , and slightly lower than those measured on an activated carbon obtained via the CO 2 -activation of a hydrochar formed from grass cuttings 19 .
Overall, the most important impact of using milk as the liquid phase in the hydrothermal carbonization to generate hydrochar-derived activated carbons for use as CO 2 sorbents was to contribute inorganic mass that adsorbed little CO 2 . This inclusion of inorganic species had the net effect of producing activated carbons that took  Table 3. CO 2 uptake capacity of some activated carbons derived from the activation of biomass or biomassderived hydrochars (HCs) under CO 2 . a Reference. www.nature.com/scientificreports www.nature.com/scientificreports/ up less CO 2 than analogous activated carbons made from hydrochars formed in water; however, the carbonaceous portions of the AC-xx-M-t and AC-xx-W-t took up similar amounts of CO 2 in adsorption processes that were energetically similar. In this way, the AC-xx-M-t behaved, at least in the context of CO 2 sorption, like composites of activated carbons and inorganics. Thus although the use of waste milk to produce hydrochar-derived activated carbons was clearly feasible, and some of these activated carbons had CO 2 uptake capacities in the same range as other activated carbons produced using CO 2 as the activation agent (1.6 mmol g −1 at 15 kPa and 0 °C), other uses of the AC-xx-M-t may be more interesting; future work will focus on applications that are favored by inorganic cations, such as calcium-catalyzed reactions.

Data availability
Figs 5 and 6 are constructed from data in Supplementary Figs S7 and S8, respectively, and Fig. 7 is constructed from data in Supplementary Figs S9−S13.