Catalytic Upgrading of a Mixed Hydroxy Acid Feedstock Derived from Kraft Black Liquor

Lignocellulosic feedstocks are widely studied for sustainable liquid fuel and chemical production. The pulp and paper industry generates large amounts of kraft black liquor (BL) from which a high volume of hydroxy acids (HAs) can be separated for further catalytic processing. Here, we explore the catalytic upgrading of HAs, including the conversion of (1) a model HA, gluconic acid; (2) a model mixture of HAs, and (3) a real mixture of HAs derived from kraft BL on M/Nb2O5 (M = Pd, Pt, Rh, and Ru). The hydrodeoxygenation of model gluconic acid reveals that “volatile” carboxylic acids (mainly C2 and C3), levulinic acid, and cyclic esters are significant products over all the catalysts, with Pd/Nb2O5 showing superior activity and selectivity toward valuable intermediates. The model mixture of HAs shows a wide range of reactivity over the supported metal catalyst, with the product selectivity strongly correlating to reaction temperature. Utilizing a 0.25% Pd/Nb2O5 catalyst, a real mixture of HAs derived from kraft BL is successfully dehydroxylated to produce a mixture rich in C3–C8 carboxylic acids that may be amenable for further upgrading, e.g., catalytically to ketones with high carbon chain lengths. Despite the feedstock complexity, we selectively cleaved the C–OH bonds of HAs, while successfully preserving most of the −COOH groups and minimizing C–C and C=O bond scission reactions under the operating conditions tested. The BL-derived HA stream is thus proposed to be a suitable platform for producing mixed carboxylic acid products from an overoxygenated byproduct feed.


Catalyst Synthesis
Synthesis of Niobium Pentoxide (Nb 2 O 5 ): The Nb 2 O 5 support was synthesized via hydrothermal method, using ammonium niobate (V) oxalate hydrate (ANO) as the precursors.In a typical synthesis, 2.7 g ANO was dissolved in 33 mL H 2 O and mechanically stirred for 5 minutes.This was followed by adding 7 mL of H 2 O 2 , and the mixture was transferred into a 60 mL Teflon-lined autoclave made of stainless steel (Parr).After sealing the reactor, the synthesis was carried out at 175 o C for 24 h followed by natural cooling.After cooling, the solid was separated from the solution using a centrifuge (ThermoFisher) washed several times with deionized water, and dried for 16 h at 80 o C. The obtained dried solid was ground using a mortar and pestle and calcined at S3 350 o C (ramp rate 2 o C/min, soak 3 h).The surface area was estimated as 144 m 2 /g (Table 1 in the main text).The supported synthesized by this method is denoted by HT-Nb 2 O 5 -175.
Synthesis of TM/Nb 2 O 5 : Supported metal catalysts (TM = Pd, Pt, Rh, and Ru) were prepared using the impregnation method.Appropriate amounts of palladium(II) nitrate hydrate, Tetraammineplatinum(II) chloride hydrate, Rhodium(III) nitrate hydrate, and ruthenium(III) nitrosyl nitrate solution (Sigma Aldrich) precursors were used to achieving a target metal weight loading of 0.25% on Nb 2 O 5 or carbon black supports.In a typical synthesis of 0.25Pd/Nb 2 O 5 , 0.0066 g Pd salt was allowed to completely dissolve in 1.4 mL H 2 O, followed by a slow addition of 0.985 g HT-Nb 2 O 5 -175 for 15 minutes, and by vigorous stirring for 5 minutes.The obtained catalysts (slurry) were dried at 80 o C for 16 h and calcined at a temperature range of 400 o C (2 o C/min ramp) for 3 h to form Pd/HT-Nb 2 O 5 -175 with nominal Pd loading of 0.25 wt%.For the carbon support, dispersion in warm deionized water (~ 60 o C) was employed before the impregnation of metal solutions.

Catalytic Reactor
Liquid-phase hydrodeoxygenation reactions were performed in an up-flow, fixed-bed stainless steel reactor containing 3 heated zones heated externally by an adequately insulated furnace (A Testing System), and the reaction temperature was monitored by an Omega-type temperature controller/thermocouple.The H 2 gas flow was controlled by a Brooks-instrument type mass flow controller while the liquid (HAs in aqueous solution) was introduced into the reactor by an Isco pump (Teledyne 500D) via a 1/8″ Swagelok tubing.Both were directed to co-currently flow upward through the catalyst bed.The product stream from the reactor outlet flowed through a highpressure gas-liquid separator.It was condensed at atmospheric temperature and collected periodically for HPLC (Supplementary HPLC analysis I) and GC (Supplementary GC analysis I) analyses.The gas separated from the liquid flowed through the back-pressure regulator (EquiliBAR ZF Series) which maintains the system pressure.Steady-state vapor samples flowing through the back-pressure regulator (BPR) were periodically collected into a sealed gas bag (Sigma-Aldrich) for offline analysis (see details below).For all experiments, the catalyst was packed in the middle of the reactor, and the remainder space was packed with quartz wool.

Feed Preparation
For experiments with gluconic acid, an 11 wt% aqueous feed was prepared by diluting commercially available 50 wt% gluconic acid (Sigma-Aldrich).The model mixed hydroxy acid feed was prepared by combining predetermined masses of formic acid, acetic acid, succinic acid, malic acid, lactic acid, glycolic acid, dimethylol propionic acid, and gluconic acid (all Sigma Aldrich) along with deionized (DI) water to achieve a total 20 wt% acid concentration in aqueous solution (Table S1).The real hydroxy acid (HA) product was separated from kraft BL [17][18] and contained 80 wt% acids.It was diluted to 8 wt% before conversion experiments.The detailed composition (31 individual acids) was determined by GC-MS analysis after derivatization, followed by GC quantification (as described in our previous work), 17 and the results are shown in Table S2.

Reactor Operation
After loading the catalyst, the temperature was raised to 300 °C (3 o C ramp rate) under H 2 flow (30 mL/min) and maintained for 2 h for reduction under 60 mL/min H 2 environment.After reduction, the reactor was cooled to the target reaction temperature and pressurized to the reaction pressure under H 2 flow.A reference pressure on the BPR was set through an N 2 -controlled flow and assumed the reactor pressure.The reference inert gas pressure was supplied to the BPR at a pressure 5 bar higher than the pressure supplied by the ISCO pump.When the reaction pressure is reached (observed by exiting H 2 flow outside the BPR), the H 2 flow rate is set to the actual flow rate for the reaction.The HA feed flow was initiated (first 0.5 mL/min) to fill the tubing before entering the reactor (specifically for 7 minutes) before switching to the target liquid flow rate (controlled by the Isco pump).The contents of the gas-liquid separator were drained after 10 minutes.The reaction product was drained through the liquid collector every 1 h, and the collected mass was weighed.

Qualitative GC/MS Analysis
The qualitative analysis of product samples was performed on a single quadrupole (Agilent 5977) GC/MSD equipped with a HP 5 MS column (30 m x 0.32 mm i.d x 0.25 μm), and a Shimadzu GCMS (QP2010) equipped with Agilent DB-FFAP column (30 m x 0.32 mm i.d x 0.25 μm) column.The use of polar and non-polar columns improves the identification of species in the product streams.The method used for analysis includes 250 °C ion source temperature, m/z between 25 and 500, and method (40 o C/1 min hold, 240 o C at 10 o C/min ramp/3 min hold).The data from the GCMS analysis were loaded onto AMDIS software containing an updated NIST database and quality identification was conducted by relying on purity level and matching percentage.The commercially available compounds identified in product streams at significant concentrations were injected into GCMS for the confirmation of retention times.The measured m/z spectra of individual compounds match the reference NIST database spectra and are depicted in Figures S3-S6.A sample labeled chromatogram of the reaction product stream, using the AMDIS/NIST reference database, is shown in Figure S7.

Supplementary HPLC Analysis
Considering that most of the HAs used in this work have low vapor pressure, GC analysis was deemed non-ideal.Thus, the quantitative analysis of HAs feeds and products were analyzed using a Shimadzu HPLC equipped with a Bio-Rad Aminex HPX87-H column (300 mm x 7.8 mm id), connected to both dual ultraviolet (UV at 210 and 250 nm) and refractive index (RID) detectors.
The HPLC method used was a mobile phase of 5mM H 2 SO 4 in deionized water, 0.5 mL/min, 50 o C oven temperature, and 180 min analysis time for each product sample injection.The injection volume used for both calibration of standard compounds and products was 5 μL, and the response factor is shown in Table S3.As the HPLC is very stable and reproducible, the external standard method of calibration was employed.The calibrated curves for each compound resulted in high linearity (R 2 value of ≥ 0.99).The HPLC peaks for the real HA feedstock and sample product (optimum) are shown in Figure S8.

Supplementary GC Analysis I
During sample analysis of the realistic HAs feed and products in the HPLC, one hydroxy acid peak overlapped with the propionic acid product peak at a retention time of 20.7 min (HPLC).Thus, the analysis of the reaction product stream was additionally conducted in an Agilent GC 7890B equipped with 2 switching valve systems connected to an Ultra-inert DB Wax column (30 m x 0.25 mm i.d x 0.25 μm) with flame ionization detector, and a PoraBOND U (25 m x 0.32 mm x 0.25 μm) column connected to a thermal conductivity detector (TCD).A typical method includes keeping the oven temperature at 35 o C for 1 min followed by 5 o C/min ramping to 180 min (held for 2 minutes) and finally 20 o C/min ramping to 250 o C (held for 2 minutes).The response factor of standard compound calibration using 1,4 dioxane as an internal standard is shown in Table S4.
The FID signal was calibrated using commercially available representative standard compounds to achieve near-perfect linear calibration curves for individual species (R 2 ≥ 0.998).The sample chromatogram of the product stream injected (1 uL injection) in GC is shown in Figure S9.

Supplementary GC Analysis II
The gas bag collected during reactions was analyzed in a 7890 Agilent gas chromatography equipped with 3 columns (MolSieve for the detection of light gases, PoraBOND U for the detection of CO 2 , and CP-Wax for the detection of alkanes/alkenes and aromatics) and three detectors (two TCDs and one FID), being controlled by 3 sampling valves for the front, back and auxiliary detectors.A rigorous calibration of more than 50 compounds consisting of light alkanes, oxygenates, CO 2, and CO was conducted, and the RF for relevant individual compounds is shown in Table S5.The sample chromatograms for the front, back, and auxiliary detectors are shown in Figures S10-S12 respectively.The HPLC and integrated GC data were then used to calculate the molar composition of each sample, the conversion, and the selectivities of products.

Catalyst Characterization
Ammonia Temperature Programmed Desorption (TPD): The acid site density of the supported catalysts was estimated using NH 3 -TPD in Autochem II 2920 (Micromeritics), equipped with a TCD detector.About 80 mg of sample was loaded into a quartz wool-packed sample tube and pretreated in a flowing 30 mL/min He at 40 o C for 30 minutes.This was followed by heating the sample to 250 o C (10 o C/min ramp rate) and reducing it at this temperature for 2 h.At 250 o C, the sample was flushed with 50 mL/min He for 30 min and finally cooled to 100 o C. Finally, 3000 ppm NH 3 /He flowing at 50 mL/min was adsorbed on the sample for 1 h.The adsorbed ammonia was removed by flushing the system with He, followed by heating the sample from 100 o C to 600 o C (10 o C/min) under He flow (50 mL/min) during which the desorbed ammonia was detected by the MS.
H 2 Pulse Chemisorption: This was conducted using an Autochem II 2920 (Micromeritics, USA) equipped with a TCD detector.About 80 mg of powdered catalyst was loaded in a U-shaped quartz tube and Ar (30 mL/min) was allowed to flow over the sample for 30 min.This was followed by reducing the sample to 10%H 2 /Ar (50 mL/min) at 400 o C for 2 h.The reduced sample was then S9 flushed with Ar at the same temperature for 1 h before cooling to 40 o C. Pulses of 10% H 2 /Ar were introduced to the sample repeatedly until the TCD peak signal remained constant between pulses.

X-ray Diffraction:
This was performed at room temperature on a PANalytical XPert PRO Alpha-1 (Malvern Panalytical) diffractometer using Cu Kα radiation in the 2θ range from 20° to 70°, at a step size of 0.0167 o .
N 2 Physisorption: A Tristar II 3020 (Micromeritics) sorption analyzer was used to collect isotherms at -196 o C using ∼100 mg of catalyst samples.Degassing under vacuum was performed at 180 °C for 3 h before measurements.

Inductively coupled plasma optical emission spectroscopy (ICP-OES):
This method was used to determine the loadings of Pd and Nb before and after the hydrodeoxygenation of the model mixture of hydroxy acids.S3) A = Peak area of species "i", C_mol i = mol of carbon in species "i"

Carbon recovery =
RF_G is the response factor derived from GC injection of standard compounds (Table S4) S10 A = Peak area of species "i", C_mol i = mol of carbon in species "i", istd = 1,4 dioxane  The commercially available organic acids in the realistic kraft-BL-derived hydroxy acids constitute ~ 52 % of the total acids and were accurately calibrated with an R 2 value ≥ 0.995.About 44 % of the acids are not commercially available for calibration.Hence, the response factors were approximated based on the class categories (# OH and # COOH) shown in Table S1.The retention times were assigned by correlating peak heights to the concentrations of these species in the mixture.Specifically, the retention times of Gluco-isosaccharinic acid (22 %), 2 hydroxybutanoic acid (7 %), 2,5 dihydroxypentanoic acid (12 %), 2-deoxypentonic acid (4 %) were easily assigned based on their relatively high concentrations in the realistic acid feedstock.An alternative estimation of conversion was designed by performing calibration using the concentration of individual acids in the mixture.A constant weight (0.1g) was added into 5 different vials containing 1 mL, 3 mL, 6 mL, 10 mL, and 15 mL DI water to obtain a calibration curve.S7.The product selectivity for the conversion of mixed model HA feed at different temperatures over 0.25%Pd/Nb 2 O 5 and 0.25%Pd/C (60 bar, 3.1 h -1 , 50 mL/min H 2 co-flow).Other compounds detected in trace amounts are acids (sorbic acid, heptanoic acid), Esters (Butyric acid 2-ethyl, hexanoic acid 2-hexenyl ester), ketones (hydroxy-2 butanone).

Figure S1 .Figure S2 .
Figure S1.Structures of the individual acids making up the mixed model hydroxy acids used in this work.

Figure S7 .Figure S8 .
Figure S7.Sample GC/MS chromatogram of product from conversion of the real HA mixture.

Figure S9 .
Figure S9.Sample GC chromatogram of reaction products from the conversion of real HA feed (Column: Ultra Inert DB Wax, Detector: FID).

Figure S17 .
Figure S17.Ammonia temperature programmed desorption profiles for bare niobia and metals supported on niobia.

Figure S25 .
Figure S25.Carbon selectivities in a catalyst-free conversion of BL-derived HA mixture at 60 bar, 0.06 mL/min, and 50 mL/min H 2 .Data were taken after 5 h on-stream.Carbon recovery: 77 %.

Figure S27 .
Figure S27.Aqueous phase product selectivities from the catalytic conversion of realistic kraft black-liquor-derived HAs over 0.25%Pd/Nb 2 O 5 at different temperatures and WHSV.These are the normalized product selectivities from Figure 5A (main text), excluding CO 2 and hydrocarbon products.The selectivity towards C 3 -C 8 carboxylic acids reached ~ 90 % (mainly C 3 -C 5 carboxylic acids).

Figure S28 .
Figure S28.XRD patterns of fresh and used 0.25Pd/Nb 2 O 5 showing the presence of different crystallographic phases of Nb 2 O 5 .The crystallite domain size approximated by the Scherrer equation revealed a slight increase from 12.2 nm (fresh) to 13.6 nm (used).

Figure S29 .Figure S30 .
Figure S29.N 2 physisorption isotherm and pore size distribution (inset) for both fresh and used 0.25Pd/Nb 2 O 5 .A decrease in the BET surface area and a slight size enlargement were observed.

Table S4 .
Response Factor from GC (Ultra inert DB Wax column).

Table S6 .
The carbon selectivity for the conversion of model gluconic acid at different Pd loadings on Nb 2 O 5 and carbon supports (230 o C, 60 bar, 2.85 h -1 , 50 mL/min H 2 co-flow).The data are not normalized.Table

Table S8 .
ICP-OES analyses of 0.25Pd/Nb 2 O 5 (fresh and used), and post-reaction mixture Product from the hydrodeoxygenation of model mixture of HAs accumulated for 10 h. *