Options for Removing Refractory Organic Substances in PreTreated Process Water from Hydrothermal Carbonization

Granular activated carbon (GAC) adsorption, as well as ozonation in combination with biodegradation was investigated in order to remove refractory organics from biologically pre-treated process waters (PW) produced by the hydrothermal carbonization (HTC) of spent grains and fine mulch. Kinetic tests revealed that the organics in spent grains PW had much lower molecular weights than organics in fine mulch PW. Moreover, isotherms showed that they were more strongly adsorbable. This was confirmed in GAC column experiments, where the breakthrough curves could be predicted fairly well by a dynamic adsorption model. On the other hand, ozonation had a stronger effect on fine mulch PW with respect to an enhancement of the aerobic degradability. Thus, the type of input material determines the properties of soluble reaction products from the carbonization process that must be accounted for when selecting the most suitable post-treatment method for HTC PW. However, adsorption on granular activated carbon should always be the final stage.


Introduction
Hydrothermal carbonization (HTC) is a process where biomass is transformed into a material comparable to brown coal.In contrast to natural processes that require many years, HTC takes less than 12 hrs.Therefore, in recent years it has been investigated as an alternative disposal option for bio waste.For the carbonized material, the so-called biochar, both an energetic utilization and an application as soil conditioner have been suggested [1].The HTC process is performed with biomass suspensions in a pressure vessel at 180 to 260 • C and resulting water vapor pressures of 15 to 60 bars [2].After the reaction phase is finished, solid products and process water (PW) are separated by filtration.The PW is acidic for most input materials.Furthermore, it contains a portion of 15% to 25% of the input organic carbon [3] resulting in chemical oxygen demand (COD) loads of 4900 to 78,700 mg/L and biochemical oxygen demand (BOD) loads of 1700 to 42,000 mg/L [2].
As summarized by Yousefifar et al. [4], the soluble reaction products comprise alcohols, ketones, aldehydes, and low molecular weight carboxylic acids.In a number of recent studies, PW has been analyzed in more detail.The results show that its main components are acetic, formic, levulinic and glycolic acid, and hydroxy-methyl-furfural [5].Moreover, 2-methoxyphenol and 2-methylbenzofuran [6], and methanol, glycerol, hydroxy acetone, and acetaldehyde [7] are also found.
Although HTC of biomass has been studied for almost ten years [8], the treatment of PW has obtained, to date, little attention.Currently, there are no technical-scale HTC plants in Europe, and the PW from pilot-plant operations is usually treated together with domestic wastewater.Kühni et al. [9] investigated the aerobic treatment of PW from carbonization of sewage sludge in continuous tests and found a COD removal efficiency of 56%, that could be increased to 70% by an upstream ultrafiltration stage.Remy et al. [10] showed, in a theoretical study, that an anaerobic stage was definitely required in order to obtain an energetic surplus when applying HTC to sewage sludge, compared to the mechanical dewatering of untreated sludge.They also estimated that an anaerobic pre-treatment stage for HTC PW followed by aerobic treatment in the activated sludge stage of a municipal sewage treatment plant together with domestic wastewater would result in an increase of the refractory COD in the effluent of the plant on the order of 17 to 25 mg/L.
Wirth and Mumme [11] and Wirth et al. [12] studied the anaerobic degradation of PW from carbonization of maize silage and sewage sludge, respectively.They found no inhibition effects; however, the test periods were rather short.Similar results were obtained by Li et al. [13] with PW from carbonization of sewage sludge.Yet, their upflow anaerobic sludge blanket (UASB) reactor was only operated for 28 days.Meanwhile, Meier et al. [14] investigated the anaerobic treatment of PW from carbonization of fine mulch together with an easily degradable co-substrate.The process was stable for about 100 days, then inhibition effects were observed that were attributed to an accumulation of organics like phenols and furfurals in the biomass.
Alternatively, Baskyr et al. [15] described the application of a wet oxidation process to PW.They obtained a COD removal of up to 30%.However, the costs of this process were quite high.As far as we know, the application of ozone to HTC PW has not been studied.It would not make sense to apply ozone to raw PW because of the high concentrations of organics and the presence of carboxylic acids and aldehydes.However, since these components are likely to be removed by biological treatment, ozonation could be a post-treatment option.Only a portion of the remaining refractory organics will consist of substances from the raw PW, the major part will probably be made up of degradation products including humic-type material.
In a preceding study, Fettig and Liebe [16] demonstrated that adsorption onto activated carbon was suitable for the removal organics from PW remaining after anaerobic and aerobic degradation.This approach was refined in our study by applying adsorption and ozonation to two different types of HTC PW, in order to develop and evaluate appropriate disposal options.
The following pre-treatment approach was applied by means of batch experiments and continuous tests on a lab-scale in this study: Nutrient recovery + anaerobic treatment + aerobic treatment [17].Taking the COD concentrations in raw PW into account, it is obvious that even when biological removal efficiencies of 90% are achieved, the concentrations of refractory COD (without dilution) will be between 500 mg/L, and some as much as 1000 mg/L.Although there are currently no regulations on the treatment of HTC PW in Germany and other European countries, one can refer to landfill leachate and other liquid residuals from waste treatment that represent a similar type of wastewater.In that case, the maximum permissible COD concentration in Germany for the direct discharge into a surface water body is 200 mg/L [18].
Therefore, this paper focuses on the removal of refractory organic substances from biologically pre-treated HTC PW by applying adsorption on granular activated carbon (GAC) and a combination of ozonation and biodegradation, respectively.

Properties of the PW
The two PW were produced by the waste management and landfill technology section of our university using a 25 L batch reactor.The conditions of the carbonization process are described in detail by Ramke et al. [19].As input materials, fine mulch, provided by a regional biomass center, and spent grains waste, provided by a local brewery, were used.
Characteristic data of the raw PW before nutrient recovery and biological pre-treatment are given in Table 1.Accordingly, the pH values were acidic while the conductivities indicate that significant amounts of salts were released from the input materials.High organic contents were reflected by COD and total organic carbon (TOC) concentrations, which were in the g/L range.BOD values were determined after 5 days (BOD 5 ) and after 30 days (BOD 30 ), respectively, with both showing the presence of both easily, and more hardly, degradable species in the PW.Meanwhile, considerable amounts of nutrients were only found in spent grains PW.

Nutrient Recovery and Biological Treatment
As a first pre-treatment step, phosphate and ammonium were recovered from spent grains PW by adding magnesium oxide (MgO) at pH 7 in order to initiate the precipitation of MgNH 4 PO 4 (struvite).Based on the nutrient concentrations given in Table 1, the magnesium dosage was calculated from the following molar ratio: Mg:NH 4 -N:PO 4 -P = 1.4:2.8:1.0.After 10 min of stirring and 2 h of sedimentation in a 10 L vessel, the residual concentrations of NH 4 -N and PO 4 -P in the supernatant were determined.Accordingly, 20.6 mmol/L (= 640 mg/L) of PO 4 -P were precipitated, but only 15.4 mmol/L (= 215 mg/L) of NH 4 -N.Thus, it can be concluded that struvite contributed to about 75% of phosphate removal while other phosphate salts with low solubility accounted for the remaining portion.The total efficiency for phosphate recovery in this stage was close to 94%, while total ammonium removal was about 25%.
Anaerobic treatment was investigated by conducting both batch tests and operating two 4 L lab reactors.In order to stabilize anaerobic degradation, beer was added as a co-substrate.For the aerobic treatment, a 7 L membrane bio-reactor (MBR) with an ultrafiltration membrane was applied.Details of these investigations are given by Meier et al. [14] and Fettig et al. [17].Since there was a dilution effect in the anaerobic stage due to co-substrate addition, it is not possible to relate the removal efficiencies directly to concentrations.Therefore, the COD removal efficiencies by anaerobic and aerobic treatment given in Table 2, were calculated from loads.Hence, the overall removal efficiencies in the biological stages were about 87% for fine mulch PW and 94.5% for spent grains PW, resulting, theoretically, in residual COD concentrations of 3125 to 4140 mg/L.Since dilution factors of 0.25 and 0.38, respectively, had to be accounted for, the actual COD concentrations of the refractory organics (see Table 2) that are relevant for the post-treatment stages, were considerably lower.

Adsorption Studies
Adsorption of pre-treated PW was studied by determining overall isotherms (dissolved organic carbon (DOC) isotherms) and by conducting kinetic tests, as well as column tests.The granular activated carbon brand Norit ROW 0.8 S (Cabot Norit, The Netherlands) was chosen for that purpose.This carbon brand is made from peat, and the median particle diameter by weight of the extruded product is 1.18 mm.The particle density is 0.642 g/cm 3 and the filter density is 0.374 g/cm 3 , corresponding to a bed porosity of 0.42.It was applied in granular form for kinetic and column tests.For the isotherm experiments, the carbon particles were crushed in a ball mill.The resulting powdered activated carbon had particle diameters below 40 µm.The isotherm data were determined by applying the bottle-point method.For each experiment, between 0.1 and 2.0 g of activated carbon were added to 0.2 L of solution.Previous tests showed that 72 h of contact time were sufficient to reach equilibrium.After equilibration, the bottles were taken from the shaker, samples were filtered through 0.45 µm membranes, and analyzed for DOC.The solid-phase DOC concentrations were calculated from a mass balance.
The isotherm data were evaluated using the ADSA software [20], which enables researchers to describe multi-component systems as a mixture of a few competing fictive species, based on the Ideal Adsorbed Solution Theory (IAST).This model, which was originally developed for gaseous mixtures and then modified for aqueous systems [21], is often used for the description of multi-component adsorption because it is solely based on single-solute parameters.
Mean film mass transfer coefficients β L for the uptake of the organics by granular activated carbon, were determined by the short fixed-bed reactor technique.From the results, effective diffusivities of the organics, as well as mean film mass transfer coefficients for other flow rates, were derived using the Gnielinski correlation.Details of this approach are described in [22].Mean intra-particle diffusivities of the organics were determined from DOC concentration-vs-time curves.For this purpose, GAC samples were placed in a basket attached to a mixing turbine that was submerged in a 2 L glass beaker, and DOC concentrations were measured as a function of time.Film mass transfer coefficients for the flow rate of about 100 m/h through the basket were estimated from the effective diffusivities.Following this, the film-homogeneous diffusion model, as outlined in [22], was fitted to the data, assuming the same value of the intra-particle diffusivity for all adsorbable components.A computer program, in dependence on the orthogonal collocation approach developed by Crittenden et al. [23], was used for the calculation.
The column experiments were conducted with two glass columns, 3 cm in diameter and 50 cm high, that were operated in series.Each column was filled with 28 cm of GAC.The PW was pumped to a constant head tank to feed the columns at a constant rate of 0.6 m/h.Thus, the total empty bed contact time was 56 min.DOC effluent concentrations were measured as a function of time.The computer model AdDesignS [24] was used to simulate the breakthrough curves of the organics in the GAC columns, based on the equilibrium and kinetic parameters derived from the batch tests.

Ozonation/Biodegradation
Ozonation was investigated in batch tests where a feed-gas stream produced from pure oxygen by an ozone generator LN 103 (Ozonia, Switzerland) was dispersed in a 2 L reactor with a glass drip that produced bubbles of about 0.2 mm in diameter.The reactor can be considered completely mixed.The feed-gas stream was 1 L/min, the feed gas concentrations were 30 to 40 mg/L of ozone, and the contact time between gas bubbles and liquid was about 10 s.The amount of ozone transferred into the liquid phase was calculated from ozone feed-gas and off-gas concentrations measured by an ozone analyzer BMT 961 (BMT Messtechnik, Stahnsdorf, Germany).
Biochemical oxygen demand after 5 days (BOD 5 ) and after 30 days (BOD 30 ) was determined in batch tests according to the manometric respirometric method using OxiTop systems (WTW, Xylem Analytics, Weilheim, Germany), using activated sludge samples from the local wastewater treatment plant.For ozonated fine mulch PW, the aerobic degradability was also evaluated by operating the membrane bioreactor described in Section 2.2.The reactor was continuously fed for 18 days with ozonated organics at a flow rate of 1 L/day.COD influent and effluent concentrations were measured, and the COD removal efficiency at steady-state was determined.
All experiments were conducted at a room temperature of 22 • C.

Analytical Methods
TOC and DOC concentrations were measured with a TOC-analyzer, Dimatoc 2000 (Dimatec, Essen, Germany) where the DOC samples were filtered through 0.45 µm membrane filters.Color and UV absorbance were measured with an UviLine 9400 spectrophotometer (Schott, Mainz, Germany).Chemical COD was determined with Merck test kits (No 1.14541 and 1.14555, respectively) after appropriate dilution of the samples.Ammonium and phosphate concentrations were measured with an Ion chromatograph 792 Basic IC (Metrohm, Herisau, Switzerland), and pH was determined with a pH meter 539 (WTW, Xylem Analytics, Weilheim, Germany).

Adsorption Studies
Overall isotherms of pre-treated HTC spent grains and fine mulch PW are shown in Figure 1.Accordingly, the non-degradable spent grains organics are very well adsorbed on activated carbon, while the solid-phase concentrations of fine mulch organics are much lower.
Water 2019, 11, x FOR PEER REVIEW 5 of 15 ozonated organics at a flow rate of 1 L/day.COD influent and effluent concentrations were measured, and the COD removal efficiency at steady-state was determined.
All experiments were conducted at a room temperature of 22 °C.

Analytical Methods
TOC and DOC concentrations were measured with a TOC-analyzer, Dimatoc 2000 (Dimatec, Essen, Germany) where the DOC samples were filtered through 0.45 µm membrane filters.Color and UV absorbance were measured with an UviLine 9400 spectrophotometer (Schott, Mainz, Germany).Chemical COD was determined with Merck test kits (No 1.14541 and 1.14555, respectively) after appropriate dilution of the samples.Ammonium and phosphate concentrations were measured with an Ion chromatograph 792 Basic IC (Metrohm, Herisau, Switzerland), and pH was determined with a pH meter 539 (WTW, Xylem Analytics, Weilheim, Germany).

Adsorption Studies
Overall isotherms of pre-treated HTC spent grains and fine mulch PW are shown in Figure 1.Accordingly, the non-degradable spent grains organics are very well adsorbed on activated carbon, while the solid-phase concentrations of fine mulch organics are much lower.The data was evaluated by fitting the IAST model to experimental data, which were obtained for two different initial concentrations (original and diluted PW).As a fitting criterion, minimization of the relative deviations between measured and calculated aqueous-phase concentrations and between experimental and calculated solid-phase concentrations was used [25].One non-adsorbable fraction, corresponding to the Freundlich coefficient K1 = 0, and four adsorbable fractions, corresponding to K2 = 10, K3 = 30, K4 = 60 and K5 = 80, respectively, were pre-specified with their Freundlich exponent being n = 0.2.Such a selection is a bit arbitrary, but it is recommended in order to limit the number of fitting parameters to the initial concentrations of the five fractions [25].Trials were also conducted with other sets of K values and with only three adsorbable components (data The data was evaluated by fitting the IAST model to experimental data, which were obtained for two different initial concentrations (original and diluted PW).As a fitting criterion, minimization of the relative deviations between measured and calculated aqueous-phase concentrations and between experimental and calculated solid-phase concentrations was used [25].One non-adsorbable fraction, corresponding to the Freundlich coefficient K 1 = 0, and four adsorbable fractions, corresponding to K 2 = 10, K 3 = 30, K 4 = 60 and K 5 = 80, respectively, were pre-specified with their Freundlich exponent being n = 0.2.Such a selection is a bit arbitrary, but it is recommended in order to limit the number of fitting parameters to the initial concentrations of the five fractions [25].Trials were also conducted with other sets of K values and with only three adsorbable components (data not shown), but significantly larger deviations were obtained.Figures 2 and 3 illustrate that the model was able to describe the adsorption equilibria quite well.
The parameters obtained by this approach are listed in Table 3.They reveal that for spent grains PW, the non-adsorbable fraction was only 1.3% and the weakly adsorbable fractions (K 2 and K 3 ) were also small, while the major portion can be characterized as strongly adsorbable (K 4 and K 5 ).The opposite result was found for fine mulch PW, with less than 20% of strongly adsorbable organics, almost 70% weakly adsorbable species, and about 12% which were considered non-adsorbable.
Table 3. Fractions of five fictive components derived from overall isotherms for pre-treated HTC process water (K 1 -K 5 and n are the Freundlich isotherm parameters of each component).
As shown in Table 4, the mean molecular weight was only 790 Dalton for PW with spent grains as an input material, and more than 10,000 Dalton for pre-treated PW from fine mulch.The latter is also much more colored (see Figure 8), presumably due to a considerable portion of non-degradable high molecular weight lignins.It can, therefore, be assumed that size exclusion effects were in part responsible for the weaker adsorbability of the organics in fine mulch PW.
Figures 4 and 5 show the data of batch tests for the determination of the intra-particle diffusivities D S .Hence, the concentration-vs-time curves can be described fairly well by the model.The mean intra-particle diffusivities derived for both PW, are also listed in Table 4.The parameters were quite similar; however, their absolute value was not significant on its own.It must rather be interpreted in the context of the kinetic model applied.Experimental and predicted breakthrough curves for column lengths of 28 cm (column 1) and 56 cm (column 2), are shown in Figures 6 and 7, respectively.The predictions were solely based on parameters determined in batch tests.The adsorption analysis data (Table 3) and kinetic parameters (Table 4) were used as input parameters for the multi-component film-homogeneous diffusion column model.The experimental data illustrate that the spent grains organics were removed quite well by GAC, where 10% breakthrough first occurred after about 75 h in the 56 cm column.In contrast, the Water 2019, 11, 730 8 of 15 fine mulch organics reached 50% breakthrough in the second column already after 20 h, hence their uptake by granular activated carbon was much more limited.
coefficients βL for a filter velocity of 0.6 m/h that were needed to model the column test data, and mean intra-particle diffusivities DS for biologically pre-treated HTC process waters.

Parameter Input Material Spent Grains
Fine Mulch DL,eff (m 2 /s) 2.6 × 10 −10 0.69 × 10 −10 MW (Dalton) 790 10,900 βL (m/s) 0.39 × 10 −5 0.14 × 10 −5 DS (m 2 /s) 6.5 × 10 −14 6.5 × 10 −14 Experimental and predicted breakthrough curves for column lengths of 28 cm (column 1) and 56 cm (column 2), are shown in Figures 6 and 7, respectively.The predictions were solely based on parameters determined in batch tests.The adsorption analysis data (Table 3) and kinetic parameters (Table 4) were used as input parameters for the multi-component film-homogeneous diffusion column model.The experimental data illustrate that the spent grains organics were removed quite well by GAC, where 10% breakthrough first occurred after about 75 h in the 56 cm column.In contrast, the fine mulch organics reached 50% breakthrough in the second column already after 20 h, hence their uptake by granular activated carbon was much more limited.Experimental and predicted breakthrough curves for column lengths of 28 cm (column 1) and 56 cm (column 2), are shown in Figures 6 and 7, respectively.The predictions were solely based on parameters determined in batch tests.The adsorption analysis data (Table 3) and kinetic parameters (Table 4) were used as input parameters for the multi-component film-homogeneous diffusion column model.The experimental data illustrate that the spent grains organics were removed quite well by GAC, where 10% breakthrough first occurred after about 75 h in the 56 cm column.In contrast, the fine mulch organics reached 50% breakthrough in the second column already after 20 h, hence their uptake by granular activated carbon was much more limited.

Ozonation Studies
The effect of ozonation on pre-treated PW using two different ozone dosages is illustrated in Figures 8 and 9.It was intended to transfer a low amount of ozone of 250 mg/L, and a higher amount of 500 mg/L to the samples.However, due to the conditions in the practical tests the actual dosages differed somewhat from these values.The ozone dosages that are assigned in Figure 9 and Table 5, respectively, were the effective amounts of ozone transferred into the liquid phase.While the lower dosages (200 to 255 mg/L) were considered reasonable for the purpose of partial oxidation, the higher dosages should show what effects a more extensive oxidation would have on the refractory organic substances.
Water 2019, 11, x FOR PEER REVIEW 9 of 15

Ozonation Studies
The effect of ozonation on pre-treated PW using two different ozone dosages is illustrated in Figures 8 and 9.It was intended to transfer a low amount of ozone of 250 mg/L, and a higher amount of 500 mg/L to the samples.However, due to the conditions in the practical tests the actual dosages differed somewhat from these values.The ozone dosages that are assigned in Figure 9 and Table 5, respectively, were the effective amounts of ozone transferred into the liquid phase.While the lower dosages (200 to 255 mg/L) were considered reasonable for the purpose of partial oxidation, the higher dosages should show what effects a more extensive oxidation would have on the refractory organic substances.The data depicted in Figure 9 show that mineralization (ΔTOC) was only on the order of 10% and COD reduction was also limited (ΔCOD = 4-36%).On the other hand, the spectral absorption coefficients (SAC) at 436 nm and 254 nm were reduced at most by 90 %to 96% and 66% to 78%, respectively, thus indicating strong changes in the molecular structure of the organic substances.As illustrated by the absolute numbers given in Table 5, UV absorbance (SAC-254) was a more sensitive parameter than color (SAC-436).The data depicted in Figure 9 show that mineralization (∆TOC) was only on the order of 10% and COD reduction was also limited (∆COD = 4-36%).On the other hand, the spectral absorption coefficients (SAC) at 436 nm and 254 nm were reduced at most by 90% to 96% and 66% to 78%, respectively, thus indicating strong changes in the molecular structure of the organic substances.As illustrated by the absolute numbers given in Table 5, UV absorbance (SAC-254) was a more sensitive parameter than color (SAC-436).
The data depicted in Figure 9 show that mineralization (ΔTOC) was only on the order of 10% and COD reduction was also limited (ΔCOD = 4-36%).On the other hand, the spectral absorption coefficients (SAC) at 436 nm and 254 nm were reduced at most by 90 %to 96% and 66% to 78%, respectively, thus indicating strong changes in the molecular structure of the organic substances.As illustrated by the absolute numbers given in Table 5, UV absorbance (SAC-254) was a more sensitive parameter than color (SAC-436).In order to quantify the effect of ozonation in more detail, the reductions of the spectral absorption coefficients were related to the ozone dosages applied.As shown in Table 5, the specific reductions were almost twice as high for ozone dosages of 200 to 255 mg/L, compared to the higher dosages.Furthermore, the reductions illustrate that ozone had a stronger impact on fine mulch organics than on spent grains organics: The specific data for the lower dosages are 5.1 compared to 3.05 m −1 /(mg O3/L) with respect to UV absorbance (ΔSAC-254) and 0.8 compared to 0.2 m −1 /(mg O3/L) with regard to decoloration (ΔSAC-436).In order to quantify the effect of ozonation in more detail, the reductions of the spectral absorption coefficients were related to the ozone dosages applied.As shown in Table 5, the specific reductions were almost twice as high for ozone dosages of 200 to 255 mg/L, compared to the higher dosages.Furthermore, the reductions that ozone had a stronger impact on fine mulch organics than on spent grains organics: The specific data for the lower dosages are 5.1 compared to 3.05 m −1 /(mg O 3 /L) with respect to UV absorbance (∆SAC-254) and 0.8 compared to 0.2 m −1 /(mg O 3 /L) with regard to decoloration (∆SAC-436).

Adsorption on Activated Carbon
A comparison of experimental and predicted breakthrough curves (Figures 6 and 7, respectively) proves that the dynamic adsorption model is suited to predict the uptake of refractory organic substances in granular activated carbon columns.This conclusion is supported by the column capacities listed in Table 6, where the average deviation between measured and predicted data was only 17%.The numbers again illustrate that the capacities for spent grains organics were much higher than for the high-molecular fine mulch organics.
As illustrated in Figure 6 it is realistic to remove > 90% of spent grains organics by granular activated carbon over a considerable period of time.Hence, this process is suited to ensure low effluent concentrations, if required.On the other hand, the removal efficiency for fine mulch organics is higher than 50% only during the first 20 h of operation.This can be attributed both to the weaker adsorbabilities of the fine mulch organics and to their slower adsorption kinetics.On a technical scale, the exact removal efficiencies and times of operation will depend on the operating conditions of the adsorber.From the data presented herein, it can be concluded that activated carbon adsorption cannot be considered an appropriate post-treatment process for fine mulch PW, while it should be considered as a promising treatment option for spent grains PW after an anaerobic and aerobic biological stage.If one assumes that column capacities of 130 mg DOC/g (≈ 390 mg COD/g) for spent grains, and 60 mg DOC/g (≈ 180 mg COD/g) for fine mulch organics, could be realized when reducing the actual COD concentrations (as shown in Table 2) to 200 mg/L by a large-scale adsorber, the carbon usage rates for the PW studied would be on the order of 1.5 and 7.6 kg/m 3 , respectively.In the preceding study [16] on the post-treatment of HTC PW, sugar beet residuals, food leftovers, and spent grains were used as input materials.The refractory organics had low to medium molecular weights, and capacities of up to 200 mg DOC/g (≈ 600 mg COD/g) were estimated for activated carbon columns.This is in line with the results obtained in the present investigation.However, more research is needed in order to optimize this post-treatment stage to also include the application of adsorbers in series.
Furthermore, one can use data from the post-treatment of landfill leachate by granular activated carbon for comparison, that has been applied on a technical scale both in Germany [26] and other countries [27,28] since the 1990s.If long contact times are ensured, column capacities of up to 350 mg COD/g have been attained for influent concentrations between 600 and 1.500 mg COD/L [26].In a pilot study, the column capacity was even 150 mg DOC/g (≈ 450 mg COD/g) for an influent concentration of 640 mg COD/L [29].After more than 25 years, the process is still used because even landfills that approach stabilization continue to produce leachate [30].These examples show that granular activated carbon is successfully applied in practice for the removal of refractory organics.

Ozonation/Biodegradation
In Table 7, the long-term BOD values (BOD 30 ) for non-ozonated and ozonated samples are listed.They reveal that the aerobic degradability was enhanced significantly by ozone with a larger effect in fine mulch PW.In both cases the lower ozone dosages seem to be sufficient in order to obtain a substantial BOD increase, as illustrated by the COD/BOD 30 -ratios.A recent study by Yang et al. [31] demonstrated that ozonation of post-hydrothermal liquefaction wastewater converted phenols into acids, and that the COD/BOD 5 -ratio was reduced to 2.44.However, in their case no biological pre-treatment was applied.Therefore, a direct comparison with our data is not possible.Ozonated fine mulch PW (ozone dosage 200 mg/L) was fed again into the membrane bioreactor, and the resulting average COD reduction was 1.8 times higher than the BOD 30 of the ozonated PW (Table 7).When calculating the ratio between ozone consumption and COD reduction for the combination of ozonation and biodegradation, the result was about 1 mg O 3 per mg of COD removed .This is in line with observations by Ried et al. [32] on the treatment of refractory organics in industrial wastewater by oxidative and biological processes.
When relating the decrease of the COD values and the formed BOD 30 values to the initial COD concentrations in Table 7, it is possible to estimate minimum removal efficiencies in an ozonation/biodegradation stage.The results are about 40% to 60% for spent grains organics and 30% to 45% for fine mulch organics.As discussed above, the actual removal efficiencies will be considerably higher.However, exact values can only be derived from long-term experiments with a membrane bioreactor.They will probably be higher than 50%, but it seems to be unlikely to reach 80% to 90% with acceptable ozone dosages.Therefore, activated carbon must also be applied here as the final treatment stage.
Although both spent grains and fine mulch organics show similar effects of ozonation, one should keep in mind that this process is quite expensive.Thus, it should not be applied for refractory organics that are strongly adsorbable, such as spent grains organics.Meanwhile, ozonation in combination with an aerobic stage, can be considered a suitable process for the removal of the high-molecular refractory organic substances from fine mulch PW.
Since no other data on ozonation of HTC PW are currently available, landfill leachate treatment is again used for comparison.In early applications in Germany, ozonation alone was used as a post-treatment process.Depending on pH, ozone dosages of 2.3 to 3.0 mg O 3 per mg COD removed were needed [33].In a technical-scale plant with varying COD concentrations, up to 2000 mg/L of ozone had to be dosed while the specific dosages were 2.2 to 2.8 mg O 3 per mg COD removed [34].In a later study, even levels of up to 3000 mg/L of ozone were required to attain a significant effect [35].Because of the high costs to produce ozone, combinations of ozonation and biological treatment were developed.As an early example, the BioQuint process rendered COD reduction of 71% to 82% possible, with specific dosages of 0.9 to 1.2 mg O 3 per mg COD removed [36].Other researchers focused on the improvement of the oxidation efficiency by raising the pH [37], adding hydrogen peroxide [37,38], or applying UV-radiation [38].However, the influence of the generation of hydroxyl radicals in these AOP processes has also been questioned with respect to leachate treatment [39].Today, it is widely acknowledged that ozonation leads to the formation of lower molecular weight products [40] that are partially biodegradable [41].Since each leachate has a unique composition, it is crucial to evaluate its behavior in oxidation processes and to select the process conditions accordingly [42].
In Figure 10 the suggested scheme for a thorough treatment of HTC PW is shown.The main sinks for organic substances are the biological stages.If nutrient recovery is an aim, this should be done upstream of the anaerobic reactor.Adding a co-substrate in the anaerobic stage can help to stabilize anaerobic degradation [14].If further removal of refractory organic matter is required, adsorption on granular activated carbon should be the first option.In case of PW with larger amounts of high-molecular weight components, this process will be quite expensive.Therefore, ozonation, in combination with aerobic degradation, should be taken into consideration instead.Here, one has the option to bring the ozonated water back to the main aerobic reactor or to use a separate reactor where the biomass can adapt more specifically to the substrate.This must be investigated further in order to find a sustainable process scheme for the treatment of that type of PW.However, adsorption on granular activated carbon should always be the final stage.
Water 2019, 11, x FOR PEER REVIEW 12 of 15 possible, with specific dosages of 0.9 to 1.2 mg O3 per mg CODremoved [36].Other researchers focused on the improvement of the oxidation efficiency by raising the pH [37], adding hydrogen peroxide [37,38], or applying UV-radiation [38].However, the influence of the generation of hydroxyl radicals in these AOP processes has also been questioned with respect to leachate treatment [39].Today, it is widely acknowledged that ozonation leads to the formation of lower molecular weight products [40] that are partially biodegradable [41].Since each leachate has a unique composition, it is crucial to evaluate its behavior in oxidation processes and to select the process conditions accordingly [42].
In Figure 10 the suggested scheme for a thorough treatment of HTC PW is shown.The main sinks for organic substances are the biological stages.If nutrient recovery is an aim, this should be done upstream of the anaerobic reactor.Adding a co-substrate in the anaerobic stage can help to stabilize anaerobic degradation [14].If further removal of refractory organic matter is required, adsorption on granular activated carbon should be the first option.In case of PW with larger amounts of high-molecular weight components, this process will be quite expensive.Therefore, ozonation, in combination with aerobic degradation, should be taken into consideration instead.Here, one has the option to bring the ozonated water back to the main aerobic reactor or to use a separate reactor where the biomass can adapt more specifically to the substrate.This must be investigated further in order to find a sustainable process scheme for the treatment of that type of PW.However, adsorption on granular activated carbon should always be the final stage.

1.
HTC PW are characterized by high concentrations of organics that require initial biological treatment.In this study the removal efficiencies of combined anaerobic-aerobic treatment were 87% for fine mulch and 94.5% for spent grains PW, respectively.2.
Adsorption onto activated carbon is well suited for the removal of the remaining refractory organics if they have low molecular weights, as in spent grains PW, because the non-adsorbable fraction is small, and the major components are quite strongly adsorbable.Thus, high column capacities can be expected.

3.
Ozonation of the refractory organics and recirculation to the aerobic stage or separate aerobic treatment should be applied if they have high molecular weights, as in fine mulch PW.In this case, COD removal can be achieved both chemically and biologically with specific ozone dosages of about 1 mg O 3 per mg of COD removed .4.
Breakthrough curves in granular activated carbon columns can be predicted by the filmhomogeneous diffusion model using the multi-component approach to describe adsorption equilibria, and mean kinetic parameters determined from batch data.Thus, a simulation tool is available for the design of technical-scale GAC columns for treatment of HTC PW. 5.
For each practical case, an economic analysis is needed in order to determine the optimal treatment train, taking the particular legal requirements into account.As proven in this study, laboratory tests can provide a basis for an evaluation of the processes that are suited for the removal of the refractory organic substances.

Figure 1 .
Figure 1.Comparison of overall isotherms of pre-treated HTC spent grains and fine mulch process water (PW).

Figure 1 .
Figure 1.Comparison of overall isotherms of pre-treated HTC spent grains and fine mulch process water (PW).

Figure 2 .
Figure 2. Overall isotherms of pre-treated HTC spent grains PW and fit of the Ideal Adsorbed Solution Theory (IAST) model.

Figure 2 .
Figure 2. Overall isotherms of pre-treated HTC spent grains PW and fit of the Ideal Adsorbed Solution Theory (IAST) model.

Figure 2 .
Figure 2. Overall isotherms of pre-treated HTC spent grains PW and fit of the Ideal Adsorbed Solution Theory (IAST) model.

Figure 3 .
Figure 3. Overall isotherms of pre-treated HTC fine mulch PW and fit of the IAST model.

2 Figure 3 .
Figure 3. Overall isotherms of pre-treated HTC fine mulch PW and fit of the IAST model.

Figure 4 .Figure 4 . 15 Figure 5 .
Figure 4. Data from a batch test with pre-treated HTC spent grains PW and fit of the filmhomogeneous diffusion model.

Figure 5 . 15 Figure 5 .
Figure 5. Data from a batch test with pre-treated HTC fine mulch PW and fit of the film-homogeneous diffusion model.

Figure 6 .
Figure 6.Experimental and predicted breakthrough curves of pre-treated HTC spent grains PW.

Figure 6 .
Figure 6.Experimental and predicted breakthrough curves of pre-treated HTC spent grains PW.

Figure 6 .
Figure 6.Experimental and predicted breakthrough curves of pre-treated HTC spent grains PW.

Figure 7 .
Figure 7. Experimental and predicted breakthrough curves of pre-treated HTC fine mulch PW.

Figure 7 .
Figure 7. Experimental and predicted breakthrough curves of pre-treated HTC fine mulch PW.

Figure 8 .
Figure 8.Effect of ozonation on pre-treated HTC spent grains PW (left) and fine mulch PW (right).

Figure 8 .
Figure 8.Effect of ozonation on pre-treated HTC spent grains PW (left) and fine mulch PW (right).

Figure 9 .
Figure 9.Effect of ozonation on pre-treated HTC spent grains PW (left) and fine mulch PW (right).

Figure 9 .
Figure 9.Effect of ozonation on pre-treated HTC spent grains PW (left) and fine mulch PW (right).

Figure 10 .
Figure 10.Suggested scheme for a thorough treatment of HTC process water.(solid lines = main treatment train; broken lines = optional stages).
PW are characterized by high concentrations of organics that require initial biological treatment.In this study the removal efficiencies of combined anaerobic-aerobic treatment were 87% for fine mulch and 94.5% for spent grains PW, respectively.

Figure 10 .
Figure 10.Suggested scheme for a thorough treatment of HTC process water.(solid lines = main treatment train; broken lines = optional stages).

Table 1 .
Analytical data of the raw hydrothermal carbonization (HTC) process waters studied.

Table 2 .
Chemical oxygen demand (COD) removal in the biological stages and residual concentrations of refractory organics.

Table 3 .
Fractions of five fictive components derived from overall isotherms for pre-treated HTC process water (K1-K5 and n are the Freundlich isotherm parameters of each component).

Table 4 .
Effective bulk diffusivities D L,eff , mean molecular weights M W , mean film mass transfer coefficients β L for a filter velocity of 0.6 m/h that were needed to model the column test data, and mean intra-particle diffusivities D S for biologically pre-treated HTC process waters.

Table 5 .
Effect of ozonation on spectral absorption coefficients of pre-treated spent grains and fine mulch PW.

Table 6 .
Comparison between measured and predicted column capacities.

Table 7 .
Effect of ozonation on the aerobic degradability of pre-treated spent grains and fine mulch PW.