Component additivity model for plastics-biomass mixtures during hydrothermal liquefaction in sub-, near-, and supercritical water.

We produced oils via hydrothermal liquefaction (HTL) of binary mixtures of biomass components (e.g., lignin, cellulose, starch) with different plastics and binary mixtures of plastics themselves. Cellulose, starch, and lignin demonstrated synergistic interactions (i.e., enhanced oil yields) with the plastics tested (polypropylene, polycarbonate, polystyrene, and polyethylene terephthalate). Polystyrene exhibited synergy during HTL with the three other plastics as did polypropylene during HTL with PET or PC. We used the experimental results to develop the first component-additivity model that predicts the oil yields from HTL of biomass-plastic and plastic-plastic mixtures. The model accounts for interactions among and between biomass components and plastic components in sub-, near-, and supercritical water. The model predicts 88% of 48 published oil yields from HTL experiments with mixtures containing plastics to within 10 wt%.

Introduction

Municipal solid waste (MSW) is about 80% carbon by mass on dry basis (United States Environmental Protection Agency, 2018), and this carbon appears in both biomass (e.g., food waste, yard waste) and plastic waste. The plastics in MSW include durable, nondurable, and packaging material, and more then two-thirds of the waste generated in each class is landfilled (United States Environmental Protection Agency, 2018). Annual MSW production in the United States has energy equivalent to 12% of annual per capita gasoline consumption. An economical, scalable, waste-to-energy process for mixed plastic-biomass wastes would both reduce landfill needs and help meet growing energy demand.

Processes such as combustion and pyrolysis can recover energy from this waste. Because MSW has a moisture content ∼20 wt% (Ozcan et al., 2016); however, there is an energy penalty (latent heat of vaporization for water) associated with such methods as they vaporize the water present (Miandad et al., 2019). Hydrothermal liquefaction (HTL) obviates this vaporization penalty by valorizing wet waste in hot, compressed water (e.g., 8–35 MPa, 280–450°C). The synthetic polymers (i.e., plastics) and natural polymers (i.e., polysaccharides, protein, lignin) in MSW decompose into products that partition, postreaction, into solid, oil, aqueous, and gas phases (Gollakota et al., 2018; Seshasayee and Savage, 2020). Energy is recovered in the oil phase. Plastics generally require supercritical HTL temperatures (Tc = 374°C) to achieve high oil yields, whereas biomass typically gives high oil yields even under subcritical temperatures (Seshasayee and Savage, 2020). To valorize mixtures of plastic and biomass at scale, one desires to operate at a single process temperature for simplicity. Hence, determining oil yields from plastic-biomass mixtures over a range of temperatures is vital for application of HTL for MSW. Previous studies (Yuan et al., 2009; Wu et al., 2017; Raikova et al., 2019; Hongthong et al., 2020; Souza Dos Passos et al., 2020; Seshasayee and Savage, 2021a, 2021b) indicate the existence of synergistic interactions between plastics and biomass that lower the depolymerization temperature of plastics and hence achieve higher oil yields for the mixtures under subcritical temperatures than would be anticipated from HTL of the components individually. Accordingly, we investigate herein the HTL of binary mixtures and then use these results to develop a model that can predict oil yields from HTL over a range of temperatures for mixtures containing synthetic and natural organic polymers.

Component additivity models predict oil yields from HTL based on the feedstock composition. We recently reported (Seshasayee and Savage, 2021a) a component additivity model that showed good predictive ability for biomass HTL over a 125°C temperature range (275–400°C). This model included biomolecule-biomolecule interactions. However, biomolecule-plastic and plastic-plastic interactions also influence HTL oil yields (Seshasayee and Savage, 2020), and these have yet to be included in a component-additivity model. In this article, we extend previous modeling efforts and present a broader component additivity model that predicts oil yields from HTL of biomass, plastics, and mixtures thereof at sub-, near-, and supercritical HTL temperatures. This model is the first of its type to include plastics and explore binary interactions between plastic-biomolecules and plastic-plastic. The model could be useful to others exploring HTL as a potential technology for producing liquid fuels from renewable biomass, waste plastics, or mixtures of the two.

Results and discussion

HTL of mixtures with plastic and biomolecule components

Previous work (Seshasayee and Savage, 2020) identified lignocelluloses as interacting with plastic mixtures (PP + PS + PET + PS) during HTL and noted that these interactions were influenced by temperature. Hence, in the present work, we performed HTL of cellulose, starch, and alkaline lignin with each plastic (PP, PS, PET, PC) at 300 (sub-), 350 (near-), and 425°C (supercritical) for 30 min.

Oil yields from HTL of single biomolecules (Seshasayee and Savage, 2020) and single plastics (Seshasayee and Savage, 2020) are used to calculate those expected for HTL of a given binary mixture in the absence of any interactions (Equation 1).

X is the mass fraction of component i in the mixture, and Y is the oil yield obtained for HTL of component i alone at the same conditions used for HTL of the mixture. If the experimental oil yield from HTL of a mixture exceeds the calculated oil yield, the interactions between the components are increasing the oil yield, i.e., there is synergy. The contrary indicates antagonism. We perform a t test with null hypothesis that the difference between the experimental and calculated values is zero and consider a difference to be statistically significant (∗∗) if p < 0.025 and statistically highly significant (∗∗∗) if p < 0.01 (Andrade, 2019). Note that the discussion of synergy herein is restricted to oil yields and does not extend to the quality of the oil produced. In prior work, we observed increases in oil yield due to synergy were accompanied with corresponding increases in energy recovery (Seshasayee and Savage, 2021b).

Figure 1 shows the experimental oil yield from HTL of each mixture along with the oil yield expected for each component, and hence the mixture, based on results from HTL of it alone at the same temperature. The results indicate statistically highly significant synergistic interactions between all three biomass components and PS during HTL at 300°C. Earlier work on HTL of PS alone showed an oil yield of 38 wt% at 300°C but much higher yield (86 wt%) at 350°C (Seshasayee and Savage, 2020, 2021b). We hypothesize that the presence of lignocelluloses in the mixtures makes this higher-temperature depolymerization process for PS available to increase oil yields at the lower HTL temperature of 300°C. PP and PET require supercritical HTL temperatures (425–450°C) to generate high oil yields (Seshasayee and Savage, 2020). Figure 1B shows a synergistic effect of the biomass similar to that for PS at 300°C for HTL of mixtures with PP and PET, albeit at a higher temperature (350°C). The greater facility of PS for forming oil during HTL alone is likely why HTL of cellulose or starch mixtures with PS at 300°C gives larger oil yield increases (experimental versus calculated) than did HTL of these components with PP or PET.

Figure 1

Experimental and calculated oil yields (wt%) from HTL of plastic + biopolymer binary mixtures

(A–C) Experimental and calculated oil yields (wt%) from HTL of binary mixtures with a biopolymer (cellulose, starch, alkaline lignin) and a plastic (PP, PS, PET, PC) at (A) 300°C, (B) 350°C, and (C) 425°C and 30 min.

For HTL at 350°C, cellulose and starch have synergistic interactions with all of the plastics tested. The increases in the experimental versus calculated oil yields for mixtures with PS are higher for HTL at 300°C than at 350°C. The smaller increase at 350°C is because PS alone can depolymerize to produce high oil yields at 350°C. There is less opportunity for the biomass to synergize oil production from PS at this condition. The increase in oil production is most likely from cellulose or starch interacting with the products from PS depolymerization. This same effect of a smaller increase in oil yield (experimental versus calculated) occurs for the same reason for HTL of PP—cellulose and PP—starch mixtures when the temperature increases from 350 to 425°C.

Alkaline lignin has synergistic interactions with PP during HTL at 425°C but not at 300 or 350°C. We hypothesize that the alkalinity of lignin offsets any PP-lignin synergy at the subcritical temperatures (Seshasayee and Savage, 2020). The ion product of water drops precipitously, however, as the temperature increases to the supercritical regime (Helmer Pedersen and Conti, 2017). The diminished ability of supercritical water to stabilize ion formation would diminish any effect of pH on HTL at 425°C and thereby allow the effects of lignin interactions with PP to become more apparent.

HTL of PC alone attains its highest oil yield at 300°C (Seshasayee and Savage, 2020) (among 300, 350, 425°C). Hence, PC has already depolymerized or liquefied at even the lowest temperature used in the present experiments. PC displays synergistic interactions with cellulose and starch during HTL at 350°C. This synergy could be due to favoring of pathways that increase oil yield from starch, cellulose, or PC. More work is needed to ascertain the source(s). Lignin shows statistically significant antagonistic behavior with PC during HTL at 300°C but not at 425°C. This binary mixture may be another that is sensitive to the large changes in pH that accompany HTL of mixtures with alkaline lignin at different temperatures.

The alkaline lignin slurry used in these experiments has a pH of 8.6 (Seshasayee and Savage, 2021a). To test whether this basic pH was influencing the experimental results, we conducted complementary HTL experiments with binary mixtures containing de-alkaline lignin. This slurry has a pH of 3.9 (Seshasayee and Savage, 2021a). We tested the effect of alkalinity on HTL at 300°C, as the ion product of water is higher at this temperature than the others investigated herein. We expect the effect of pH, if any, to be more pronounced at this HTL temperature.

Figure 2 shows the oil yields from HTL of mixtures with alkaline or dealkaline lignin are about the same for PP, PET, and PS. The mixtures with PET and PP show no statistically significant synergy or antagonism. The mixtures with PS both show statistically significant synergy. HTL of both lignins with PC show antagonism, but the effect is large enough to be statistically significant only from alkaline lignin. This result supports the hypothesis above that the alkalinity in lignin is causing the reduced oil yield.

Figure 2

Experimental and calculated oil yields (wt%) from HTL of binary mixtures with alkaline or de-alkaline lignin and PP, PS, PET, or PC at 300°C and 30 min

HTL of mixtures of plastics

Our previous work (Seshasayee and Savage, 2020) identified synergistic interactions for HTL of an equi-mass mixture of PP, PS, PET, and PC at 300 and 400°C. During HTL at 350 and 425°C, the interactions turned antagonistic. To explore plastic-plastic interactions more fully, we performed HTL of binary mixtures of these plastics at 300, 350, and 425°C. Figure 3 displays the results.

Figure 3

Experimental and calculated oil yields (wt%) from HTL of plastic + plastic binary mixtures

(A–C) Experimental and calculated oil yields (wt%) from HTL of binary mixtures (plastic A + plastic B) with PP, PS, PET, or PC at (A) 300°C, (B) 350°C, and (C) 425°C and 30 min.

As was the case with the biomass components, HTL at 300°C of the plastics shows synergy with PS (Figure 3A). All of the binary mixtures with PP or PS show synergistic interactions during HTL at 425°C, whereas all of the binary mixtures with PP show antagonistic behavior during HTL at 350°C. Except for the results from HTL at 425°C, the observations in Figure 3 are consistent with the previous report on synergy or antagonism during HTL of an equi-mass plastic mixtures (Seshasayee and Savage, 2020).

The liquefaction temperature (as judged by oil formation) follows the order PC < PS < PP < PET for HTL of a plastic alone (Seshasayee and Savage, 2020). Hence, for HTL of a mixture, reactive intermediates from PC could interact with PS and facilitate its liquefaction, which in turn could further accelerate liquefaction of PP and PET. Thus, HTL of a multicomponent mixture (e.g., with four plastics) could produce more oil at a lower temperature than would a single component or binary mixture. Note that HTL at 425°C produces a larger increase in oil yield (experimental versus calculated) for the PP-PC mixture than for the PP-PS and PP-PET binaries. This result is consistent with the aforementioned hypothesis that intermediates from components liquefying at lower temperature can lead to greater synergy.

Component additivity model

Component additivity models are used to determine oil yields from HTL. They include unary component terms, which account for oil production during HTL of a component alone, and can also include binary interaction factors (Li et al., 2017; Lu et al., 2018; Yang et al., 2019). In previous work (Seshasayee and Savage, 2021a), we reported a component-additivity model (see Equation 2) that determined oil yields (Yoil) from HTL of biomass components at 275–400°C and had better predictive ability than previous component-additivity models.

Here, we extend this model to include plastics, plastic-biomolecule interactions, and plastic-plastic interactions. The ai and aij coefficients in the model (given in Figure 4) are determined using experimental data from HTL of individual components (Seshasayee and Savage, 2020) and binary mixtures (see Figures 1, 3 and Seshasayee and Savage, 2021a). The coefficients have different values at the three different HTL temperatures examined herein.

Figure 4

Coefficients in Equation 2 determined at different HTL temperatures

Interactions are highlighted in shades of green and red to indicate degrees of synergy and antagonism, respectively.

The component additivity model employs just one parameter (a) for each component and one parameter for each binary combination (a). This amount is the minimum number of parameters needed to predict oil yields from HTL of mixtures with plastics. The a parameter is simply the oil yield observed experimentally from HTL of that component alone at a given HTL time and temperature (i.e., the Y values in Equation 1). The a parameter for each binary mixture is determined directly from an HTL experiment with each binary mixture (see supplemental information Methods S1 for more details). The uncertainties of each of the parameters in the model are given in Table S1.

In this section, we evaluate the efficacy of the model for predicting published oil yields from HTL of mixtures containing plastics. These are true predictions (not an assessment of goodness of fit) in that none of these data were used to determine the model parameters. We found 48 experimental oil yields published for HTL (t ≥ 15 min) of mixtures containing PP, PET, PS, and/or PC with and/or without biomass (Wu et al., 2017; Raikova et al., 2019; Souza Dos Passos et al., 2020; Seshasayee and Savage, 2021b). One of the studies (Raikova et al., 2019) was not done isothermally but rather used a fixed heating rate to the desired final temperature and then immediate cooling. Given the scarcity of literature reports outside our lab on oil yields from HTL of mixtures with these plastics, we opted to include this nonisothermal study in assessing model performance. The 48 published oil yields were predicted using Equation 2 with the coefficients in Figure 4. We used the coefficients from HTL at 300°C to predict literature oil yields for HTL between 280 and 325°C. Similarly, coefficients for HTL at 350 and 425°C were used to predict oil yields from published studies of HTL between 326–400°C and 401–425°C, respectively.

Figure 5 provides a parity plot that compares the predicted and experimental oil yields, and Table S2 provides the details. We assess the accuracy of the model using two metrics—the percentage of experimental oil yields predicted to within 10 wt% and the average absolute error. For the first metric, 88% of the predicted oil yields (42 of 48) were within 10 wt% of the experimental value and 59% (24 of 48) were within 5 wt% of the experimental oil yields. The average absolute error in the predicted oil yield is 5.0 wt%. This predictive ability is better than that of prior component additivity models, all of which were parameterized solely for HTL of biomass (Seshasayee and Savage, 2021a). This model is the first to be also parameterized for HTL of plastic components.

Figure 5

Comparison of experimental and predicted (Equation 2) oil yields (wt%) for HTL of mixtures with plastics

Red dotted line represents 5 wt% deviation. Green dotted line represents 10 wt% deviation. Table S2 provides the data and details.

Four of the published oil yields that the model fails to predict to within ±10 wt% are overpredicted. The two points at the highest oil yields are from our prior work, and they are the only two cases where the mixture had greater than 6 wt% lipid. The lipid content was 78% for these experiments. We have not yet tested for potential interactions between lipids and the plastics used herein, but that would be useful for future work. Another oil yield that was greatly overpredicted is from a recent study from Biller's lab (Souza Dos Passos et al., 2020). This mixture was 50 wt% PS, which gave an oil yield of 86 wt% at 350°C in our earlier experimental work (Seshasayee and Savage, 2020). Even if the biomass in the other half of the mixture contributed nothing to the oil yield, we would expect an oil yield of 43 wt%. The experimental oil yield reported for the mixture was just 16 wt%. We used DCM to recover oil after HTL, whereas the Biller study (Souza Dos Passos et al., 2020) used methanol. To determine whether the different solvents were causing the widely different oil yields, we ran an experiment for HTL of PS at 425°C, 30 min with methanol as the recovery solvent (following the recovery method of the Biller study (Souza Dos Passos et al., 2020)) and measured an oil yield of 13.32 ± 0.72 wt%. An analogous experiment with DCM as recovery solvent (following the recovery method of the Biller study (Souza Dos Passos et al., 2020)) for HTL of PS at 425°C, 30 min gave an oil yield of 47.03 ± 0.13 wt%. The reason for this difference is the different solvents used to recover oil in the present work and in the Biller study. Thus, understanding the influence of solvent on dissolution of the post-HTL reactor contents requires additional work.

Conclusions

The component additivity model reported herein predicts well the oil yields from HTL of mixtures with plastics. Eighty-eight percent of 48 published oil yields were predicted to within 10 wt% of the experimental value, and 50% of 48 data points were predicted within 5 wt%. The average absolute error in the predicted oil yield was 5.0 wt%. This model can be used to predict oil yields from isothermal HTL of mixtures with biomass (e.g., microalgae, sludges, food waste, agricultural residues) and plastics and thus better enables preliminary process design work and technoeconomic analyses for HTL of a variety of renewable biomass and waste plastic feedstocks.

There are 11 types of plastics widely used in the consumer market (Geyer et al., 2017). The present work focuses on just four (PS, PET, PC, PP). Experimental work with a larger set of plastics would permit expansion of the model and further broaden its utility.

A given binary mixture with a plastic component can exhibit synergy at one HTL temperature and no effect or antagonism at another. We hypothesize that this influence of temperature is linked, at least in part, to the liquefaction behavior of the plastic alone. That is, synergy can be observed when the HTL temperature is 50°C or so lower than the HTL temperature where the plastic alone gives high oil yields. Examples of this behavior are the significant synergies for HTL of PP with cellulose or starch at 350°C and HTL of PS with cellulose or starch at 300°C. These temperatures are about 50°C less than those where the synthetic polymers would give high oil yields from HTL alone. We hypothesize that addition of lignocelluloses expedites the liquefaction process for the synthetic polymer and thereby produces oil from the plastic at temperatures lower than would be expected based on HTL of the plastic alone.

Limitations of the study

The authors acknowledge the following to be limitations of the study:

The predictive ability of the model presented herein was tested against 48 data points we found in the literature. This limited availability of literature data for HTL with plastics limits the extent to which the model can be tested. Larger datasets and additional experimental results would be helpful in assessing the predictive ability of the model with greater confidence.

We used dichloromethane to dissolve the oil after the HTL experiments, and these results were used to parameterize the model. The use of a different solvent in HTL experiments could solubilize a different portion of the molecules in the reactor after HTL and give a different oil yield. The effect of solvent on HTL of plastics has not been explored extensively, and the present model is best applied to studies wherein dichloromethane was used to recover the oil after HTL.

The present model addresses only oil yields and does not predict other important HTL outcomes such as the higher heating value and overall energy recovery in the oil. Additional work is needed to characterize the oils produced from HTL of plastics and biomass mixtures so these and additional outcomes could be predicted.

STAR★Methods

Key resources table

Resource availability

Lead contact

Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Phillip Savage (pes15@psu.edu).

Materials availability

Cellulose (microcrystalline), potato starch, alkaline lignin, de-alkaline lignin, polypropylene (PP), polycarbonate (PC), polystyrene (PS), and polyethylene terephthalate (PET) were all obtained commercially and used as received. Reagent-grade dichloromethane (DCM) was procured from VWR International, while deionized water was produced in-house. We constructed 10 mL (internal volume) 316 stainless steel mini-batch reactors from ¾ in. Swagelok parts (a port connector and two caps).

Method details

We loaded 0.3892 g of feedstock into the mini-batch reactors. The mass ratio was 78:22 for biopolymer – plastic mixtures, which reflects their proportions in MSW, and 50:50 for binary mixtures of plastics. Sufficient water was added to achieve a pressure of 25 MPa for HTL at 425°C and saturation pressure for HTL at 300 and 350°C. The sealed reactors were immersed in an isothermal fluidized sand bath (Techne) at the desired HTL temperature for 30 min. Triplicate runs at each condition allowed for calculation of the mean oil yield and standard deviation.

After the reaction, the contents of the cooled reactor were recovered using 20 mL of DCM and 20 mL of deionized water, used in smaller aliquots. Some dissolution of any unreacted plastics may occur in DCM, and these amounts have been reported previously (Seshasayee and Savage, 2020). Solid products were removed in a 1 μm glass fiber syringe filter from Tisch Scientific (North Bend, OH). The DCM-water mixture was centrifuged for 8 min at 3000 RPM and each phase was recovered and then dried. The material remaining after DCM evaporation is the oil and its yield (Yoil) is determined gravimetrically (Equation 3).

Quantification and statistical analysis

All experiments are run in triplicates to obtain the mean yield and its standard deviation. For comparisons of experimental and calculated oil yields, we perform a t test with null hypothesis that the difference between the experimental and calculated values is zero and consider a difference to be statistically significant (∗∗) if p < 0.025 and statistically highly significant (∗∗∗) if p < 0.01.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant proteins
Cellulose (microcrystalline)	TCI America	A17730-36
Potato starch	TCI America	A11961-36
Alkaline lignin	TCI America	L0082
De-alkaline lignin	TCI America	L0045
Polypropylene	Sigma Aldrich	427888
Polystyrene	Sigma Aldrich	331651
Polycarbonate	Sigma Aldrich	181625
Polyethylene Terephthalate	Sigma Aldrich	429252
Software and algorithms
For t-test	https://www.medcalc.org/calc/comparison_of_means.php