Design and cold flow testing of a Gas-Solid Vortex Reactor demonstration unit for biomass fast pyrolysis

Accepted Manuscript Design and Cold Flow Testing of a Gas-Solid Vortex Reactor Demonstration Unit for Biomass Fast Pyrolysis Arturo Gonzalez-Quiroga, Pieter A. Reyniers, Shekhar R. Kulkarni, Maria M. Torregrosa, Patrice Perreault, Geraldine J. Heynderickx, Kevin M. Van Geem, Guy B. Marin PII: DOI: Reference:

S1385-8947(17)30954-3 http://dx.doi.org/10.1016/j.cej.2017.06.003 CEJ 17084

To appear in:

Chemical Engineering Journal

Received Date: Revised Date: Accepted Date:

16 January 2017 22 May 2017 1 June 2017

Please cite this article as: A. Gonzalez-Quiroga, P.A. Reyniers, S.R. Kulkarni, M.M. Torregrosa, P. Perreault, G.J. Heynderickx, K.M. Van Geem, G.B. Marin, Design and Cold Flow Testing of a Gas-Solid Vortex Reactor Demonstration Unit for Biomass Fast Pyrolysis, Chemical Engineering Journal (2017), doi: http://dx.doi.org/ 10.1016/j.cej.2017.06.003

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Design and Cold Flow Testing of a Gas-Solid Vortex Reactor Demonstration Unit for Biomass Fast Pyrolysis Arturo Gonzalez-Quiroga, Pieter A. Reyniers, Shekhar R. Kulkarni, Maria M. Torregrosa, Patrice Perreault, Geraldine J. Heynderickx, Kevin M. Van Geem, Guy B. Marin Laboratory for Chemical Technology, Ghent University, Technologiepark-Zwijnaarde 914 - 9052 Ghent, Belgium. Abstract Innovative gas-solid fluidized beds with process intensification capabilities are among the most promising alternatives for the current state of the art in the chemical industry. In the present work the advantages of such a reactor that sustains a rotating fluidized bed with gas-solid slip velocities much higher than those in conventional fluidized beds are illustrated computationally and experimentally. A Gas-Solid Vortex Reactor (GSVR) demonstration unit is designed to operate at typical biomass fast pyrolysis conditions targeting the production of chemicals and fuels from renewable feedstocks. For the demonstration unit preheated N2 supplies the thermal energy required by the fast pyrolysis process but alternative sources can also be evaluated. A broad range of operation conditions in the 80 mm diameter and 15 mm height GSVR can be evaluated: N2 mass flow rates of 5-10 g s-1 and biomass feed mass flow rates of 0.14-1.4 g s-1. Particle-free and particulate flow experiments confirmed that the carrier gas is evenly distributed around the GSVR cylindrical chamber as anticipated by computational fluid dynamic simulations. The latter also supported the inclusion of a profiled bottom end wall and a diverging exhaust. Cold flow experiments with biomass confirmed that the GSVR sustains a rotating fluidized bed with average bed height of 10 mm and solids azimuthal velocities of 6-7 m s-1. Keywords: reactors; fluidized bed; process intensification; biomass; fast pyrolysis; bio-oil.

1

1. Introduction The diversification of natural resources for the sustainable production of chemicals and fuels and for the conversion of energy requires the development of innovative reactor technologies. Ideally, orders of magnitude reduction in size, improved control and enhanced heat, mass and momentum transfer are combined with a safe, cost-effective and energy-efficient operation. This has been generally referred to as process intensification [1]. Gas-solid fluidized beds (FBs), widely used for reactive and nonreactive processes in the chemical industry, are the focus in this work [2]. We present the design and cold flow testing of a Gas-Solid Vortex Reactor (GSVR) demonstration unit that enables fluidization in a centrifugal field [3-5]. A dense fluidized bed with bed width-to-height ratio and gas-solid slip velocity much higher than those in gravitational fluidized beds can be sustained [6-9]. Higher gassolid slip velocity leads to intensified interfacial transfer of heat, mass and momentum. Many industrial processes that rely on gas-solid contact can be implemented in the GSVR; in this work, the transformation of biomass via fast pyrolysis is addressed. The latter is regarded as one of the key potential technologies for biomass valorization [10]. The fast pyrolysis bio-oil is an extremely complex mixture of aromatic and nonaromatic oxygenates, e.g., alcohols, carboxylic acids, aldehydes, esters, ketones, furans, pyrans, carbohydrates as well as large molecular oligomers and nitrogen containing compounds [11, 12]. The GSVR technology can potentially benefit the biomass fast pyrolysis process in terms of both bio-oil yield and bio-oil quality. Conventional FBs are limited to gas-solid slip velocities that do not exceed the terminal velocity of the solid particles in the earth gravitational field [13]. Additional drawbacks of gravitational FBs are the decrease in bed density with increasing gas velocity and the non-uniformity of the bed [2]. In the GSVR, gas is injected at high velocity via tangentially oriented inlet slots in a cylindrical chamber in which solids are continuously fed [4, 5, 13]. Momentum transfers from the gas to the solids, causing the latter to rotate, thus generating a large radially outward centrifugal force which opposes the radially inward gas-solid drag force. A sufficiently high centrifugal-to-drag force ratio in the GSVR removes the limitation in gas-solid slip velocity and provides the opportunity for significantly increasing the efficiency. Moreover, both the drag and the counteracting centrifugal force increase 2

roughly equally with increasing gas flow rate within a wide gas flow range [13]. Therefore, a GSVR of given design offers a high flexibility with respect to the gas flow rate. Both experimental and numerical studies point out the process intensification potential of the GSVR. Several industrially relevant processes have been suggested for implementation in the GSVR: fluid catalytic cracking (FCC) [14], coating of cohesive particles [15], gas adsorption [16], drying [17], gasification [18], combustion [19] and pyrolysis [18, 20, 21]. However, to the best of the authors’ knowledge the current unit is the first reactive GSVR actually being constructed. Several fast pyrolysis reactor configurations have been developed and some technologies have been scaled up to demonstration scale. The most relevant are spouted bed [22], static and circulating fluidized beds (SFB and CFB, respectively) [23, 24], rotating cone [25, 26], auger [27], and ablative [28] reactors. The latter category also includes the so-called vortex pyrolysis reactor [29] developed in the 90’s at the National Renewable Energy Laboratory (NREL) which, apart from the name, differs substantially from the GSVR studied in this work. With the currently established reactor technologies it has not been possible to reconcile the actual and the ideal operation conditions, i.e., high interfacial heat transfer, rapid removal of the pyrolysis vapors, and precise temperature control [30, 31]. In the GSVR, convective heat transfer coefficients that are three to five times higher than those in conventional FBs can be reached [20]. The estimated residence time of the pyrolysis vapors before reaching the quenching section ranges from 50 to 110 ms. These residence times are substantially lower than those in the other fast pyrolysis reactors mentioned above in which they vary from 0.5 to 2 s. The enhanced heat transfer and bed uniformity allows to gain improved control on the pyrolysis temperature. Computational fluid dynamics (CFD) simulations of biomass fast pyrolysis in a GSVR show that the gas and solid phases rapidly reach thermal equilibrium after entering the reactor chamber, approaching a final temperature that can be adjusted via the gas-to-biomass mass flow ratio [20]. As a consequence of the improved temperature control and bed uniformity, it is possible to produce bio-oils with a higher selectivity towards targeted components. By combining particle-free CFD simulations with particle-free and particulate flow experiments in a new GSVR, this work represents the next step in the development of this technology. First, a detailed 3

description of the GSVR demonstration unit for biomass fast pyrolysis is presented. Additionally, the fast pyrolysis regime in the GSVR is compared to that in conventional FBs. The sizing of the GSVR is then addressed based on steady state mass and energy balances together with estimations of the centrifugal-to-drag force ratio. Subsequently, particle-free CFD simulations are used to assess three new GSVR design features: a single main gas inlet, a profiled bottom end wall and a diverging exhaust. Finally, the cold flow testing for both particle-free and particulate flow is presented. 2. Computational and experimental procedures In total, 11 GSVR designs were tested via Reynolds-averaged Navier-Stokes (RANS) simulations of single-phase N2 flow using the commercial simulation package ANSYS Fluent® 15.0.7. The gas mass flow rate equaled 6.7 g s-1 at an inlet pressure of 106 kPa and inlet temperature of 842 K. The effects of turbulence were accounted for via the Reynolds stress model (RSM) with linear pressure-strain. The quadratic upstream interpolation for convective kinematics (QUICK) discretization scheme was used for all equations, except for the pressure equation which was discretized using a second order central scheme. In the cold flow experiments, an absolute pressure sensor with a span of 80-160 kPa and 11 milliampere output differential pressure sensors (Unik 5000) with a response frequency of 3.5 kHz and an accuracy of ±0.04% of full scale, were used. Radial profiles of solids azimuthal velocity were measured with a 2D standard Particle Image Velocimetry (PIV) setup from LaVision® as shown in Figure 1. The bottom end wall of the GSVR consisted of a non-profiled transparent polycarbonate glass to allow proper visualization of the rotating fluidized bed. For particulate flow experiments the unit operated in a semi-batch mode, i.e., carrier gas was continuously fed while the solids feeding was stopped for PIV measurements. Solids consisted of pinewood with an average particle density of 500 kg m-3 and maximum dimension of 1.5 mm.

4

Figure 1. Schematics of the experimental configuration for PIV data acquisition, showing a) half of the GSVR bottom end wall with the highlighted PIV test section and b) side view of the GSVR. A set of optics is used to direct the diffused laser light towards the test section, and the PIV camera is set perpendicular to the bottom end wall to take images of the illuminated particles from the fluidized bed. The PIV setup was equipped with a 4 MP CCD camera (ImagerProX4M) and a 135 mJ, 15 Hz, Nd:YAG Litron laser. PIV is widely used to measure gas velocities via illuminated small tracers particles (<20 μm) in a plane. In the present work, the 2D PIV is used to measure the azimuthal velocity of biomass particles of 1.5 mm maximum dimension based on camera images having 10-40 pixels per particle. Further details on the application of the PIV technique are given by Kovacevic et al. [7]. 3. Design of the GSVR demonstration unit In what follows the five main sections of the GSVR demonstration unit are described in detail. GSVR, SFB and CFB are then compared in terms of their fast pyrolysis operation regimes. Biomass and 5

carrier gas mass flow rates have been calculated from steady-state mass and energy balances. These mass flow rate ranges are used, together with centrifugal-to-drag force ratio estimations, to define the dimensions of the GSVR. Finally, particle-free CFD simulations support the introduction of three novel design features; namely, the single main gas inlet, the profiled bottom end wall and the diverging exhaust. 3.1. Unit description and pyrolysis regime in the GSVR The GSVR has been designed with the objective to have a proof-of-concept unit that allows to study high temperature (up to ~1200 K) reactive processes involving biomass. Temperature and pressure control, along with solids feeding accuracy (SFA) and data collection, are key drivers of the design. The unit consists of five main sections as shown in the block diagram in Figure 2: biomass feeding, N2 supply and conditioning, GSVR, solids separation and bio-oil condensation.

Figure 2. Simplified block diagram of the Gas-Solid Vortex Reactor demonstration unit for biomass fast pyrolysis. A more detailed description of the unit is shown in the process flow diagram in Figure 3. A gravimetric feeder (model KMLSFSKT20, Coperion K-Tron) delivers biomass into a custom made 10 mm diameter injector screw which conveys the solids to the GSVR chamber. The gravimetric feeder has two interchangeable sets of fine pitch and coarse pitch auger screws which allow to handle biomass mass flow rates from 0.10 to 1.7 g s-1. The SFA is quantified as the relative standard deviation 6

based on samples taken over 60 s intervals and should at least be within 5% [32]. In the present work, however, an upper SFA limit of 2.5% has been adopted. The SFA is negatively affected when the biomass mass flow rate decreases but is not altered by the injector screw provided that there is no solids accumulation. Tests with grinded poplar of 1.5 mm maximum dimension, a fibrous material with a high arching tendency, showed that a SFA of 2.5% corresponds to a minimum biomass mass flow rate of 0.14 g s-1. This is a rather conservative value for biomass particles with lower arching tendency. For example, the solids feeding accuracy for pinewood particles of 1.5 mm maximum dimension at this mass flow rate was 1.5%. The arching tendency is mainly determined by particle shape, mean size, size distribution, moisture content, compressibility and surface characteristics [32].

Figure 3. Process flow diagram of the Gas-Solid Vortex Reactor demonstration unit for biomass fast pyrolysis. The air feeding line is used for cold flow experiments. temperature transmitter,

pressure indicator,

represents temperature indicator,

pressure transmitter and

flow transmitter.

The combination gravimetric feeder-injector screw has been tested with success in other fast pyrolysis setups [33]; however, design parameters for the injector screw are not reported. A custom designed 7

injector screw with constant-pitch and 80% void space suffices for most of the biomass types but is limited to approximately 0.8 g s-1 for solids with a high arching tendency. A leak-tight metallic enclosure connected to a N2 bottle surrounds the gravimetric feeder. This allows to control the pressure on top of the hopper up to a maximum allowable pressure of 300 kPa and to establish an inert atmosphere, thus avoiding oxygen from air entering the unit. Coolant circulates through the jacket of the injector screw (thermostat Alpha RA8, Lauda) to prevent reactions in the solids feeding line and to minimize the risk for blockages. Remark that controlled feeding of biomass into reactors operating at high temperature and even slight overpressure has been recognized as a challenging task [32, 34]. Liquid N2 is available in containers of 100 kg capacity which are convenient for handling and refilling. The maximum N2 mass flow rate that can be set is 10 g s-1, equivalent to a 104 s run. For this demonstration unit preheated N2 supplies the thermal energy required by the fast pyrolysis process. This thermal energy exchange suffices for an appropriate range of biomass flows, as it is shown in section 3.2. A heat carrier loop (HCL) expands the range of biomass flow rates and can be implemented in future development stages. Also alternative gasses can be used such as hydrogen or even recycled gas. This is substantially different than the classically applied HCLs for fast pyrolysis reactors, which consist of a stream of sand or steel balls in thermal contact with flue gases from char combustion [23, 35]. The maximum N2 inlet temperature is set at 930 K to minimize gas-phase cracking reactions of the pyrolysis vapors at the entrance of the bed. For modeling purposes, the outlet temperature of the solid phase is fixed at 773 K. This pyrolysis temperature corresponds to the highest bio-oil yield for several woody biomasses [36, 37]. Liquid N2 flows from the 100 kg container into a 5 kW electric vaporizer (model SEB 1R2T125380, Cryoquip Ltd) which delivers N2 at 278 K and up to 500 kPa. Subsequently, N2 flows through an electric heater (Kanthal flow heater, Sandvik) in which the gas temperature is increased to within the range 773-930 K. The required electric power varies from 3 to 7 kW.

8

The solids separation section consists of a 50 mm diameter high-throughput cyclone in series with an 80 mm diameter high-efficiency cyclone. The dimensions of the cyclones result from scaling performance curves according to the Stairmand method [38]. Both cyclones and their connection lines incorporate electric heating to maintain the gas temperature at 773 K until it reaches the condensation section. The gas volumetric flow rate in the solid separation section, at pyrolysis conditions, varies from 7.5 10-3 to 17 10-3 m3 s-1. At these conditions, the calculated maximum pressure drop in both cyclones amounts to 7 kPa. The solids separation efficiency of the cyclone decreases as both the density and size of the solids decrease. For the lowest gas flow rate and a solid density of 250 kg m-3 the calculated separation efficiency for 30 μm fines in the high-throughput cyclone is 90%. At the same operation conditions, the separation efficiency in the high-efficiency cyclone is 96%. Bio-oil condenses in a double pipe heat exchanger folded as a double U. This arrangement allows to collect a carbon-rich and a water-rich bio-oil fraction. Gas exiting the solids separation section enters tangentially into the first U. This increases its average residence time and enhances the inner-tube heat transfer coefficient. Water at a maximum volumetric flow rate of 1.0 10-4 m3 s-1 flows inside the jacket of the first U providing cooling rates of up to 7 kW. A silicon oil thermostat (model Integral XT 550, Lauda) connected to the jacket of the second U provides cooling rates up to 1.8 kW at 243 K. The biooil recovery incorporates a single-stage electrostatic precipitator [39] and a wire demister. The former consists of two SS316, 2 mm diameter and 1 m length, electrodes connected to a 15 kV power supply (model SPL-I-AC-15N50, HVP GmbH). Simulations carried out in ASPEN Plus® V9 show the effects of gas outlet temperature on bio-oil recovery and cooling rate. Pyrolysis gas and bio-oil molecular compositions from literature [10, 40] (see Table S1 in Supplementary Material) complement the global compositions from the mass and energy balances in section 3.2. Bio-oil recovery ranges from 95 to 99 wt. % for N2-to-biomass mass flow ratios from 36 to 9 at an outlet temperature of 248 K. A global heat transfer coefficient of 100 W m-2 K-1 indicates a heat transfer surface area of 1.2 m2. This heat transfer coefficient is at the lower limit of the recommended values for refrigerated volatile organic compounds (VOC) condensers [41].

9

The internal pipe diameter is 10 cm, the jacketed length is 4 m and the annular space is 1 cm. The latter includes an induced spiral path to enhance the heat transfer coefficient. The pyrolysis (

) and Biot (

) dimensionless groups reveal important differences between heat

transfer and reaction time scales in the GSVR and those in CFBs and SFBs [42]. Equation 1 and represents the inverse of the first Damköhler number intraparticle conduction time scales. second Damköhler number Finally,

is given in

, i.e., the ratio of reaction and

is given in Equation 2 and is equal to the inverse of the

, i.e., the ratio of reaction and interfacial convection time scales.

is given in Equation 3 and represents the ratio of intraparticle conduction and interfacial

convection time scales. Interfacial convection is the dominant heat transfer mechanism in CFBs, SFBs and GSVR. The contribution of particle-particle relative to gas-particle heat transfer in FBs has been assessed via CFD simulations. Chang et al. [43] found that for dense FBs of a binary mixture the ratio of particle-particle heat transfer to gas-particle heat transfer ranges from 0.013 to 0.031. Shu et al. [44] found that the for downer reactors the effect of particle-particle heat transfer is negligible compared to the effect of gassolid particle heat transfer. In other words, excluding radiative heat transfer, heat transfer between cold and hot particles occurs predominantly in an indirect way. Thermal energy is first transferred from hot particles to gas, and then transferred from gas to cold particles. CFD simulations show that radiative heat transfer at the operation conditions of the GSVR with biomass pyrolysis accounts for approximately 10% of the total volumetric heat transfer [20]. Even without a solid heat carrier in the GSVR, a comparison between fluidized bed reactor technologies based on the convective heat transfer coefficient should be made. (1)

(2)

(3)

10

The thermal conductivity, solid density and biomass heat capacity are represented by respectively (

0.11 W m-1 K-1,

characteristic length scale, the carrier gas and

500 kg m-3 and

1450 J kg-1 K-1).

,

and

,

represents the

the convective heat transfer coefficient between the biomass particle and

the pyrolysis reaction rate coefficient. The convective heat transfer coefficient in

the GSVR is determined via the Gunn correlation [45]. Biomass particles are considered as cylinders with aspect ratios of 2.5 and a maximum dimension of 2.5 mm [46]. The characteristic length for heat transfer is the diameter of the particle [47]. The pyrolysis reaction rate coefficient of the initial decomposition of cellulose is taken as representative of biomass pyrolysis [20, 48]. The convective heat transfer coefficient in FBs and CFBs also include wall-to-bed heat transfer [49]. Plotting characteristic length vs. the convective heat transfer coefficient results in the different pyrolysis regimes delimited by the Pyrolysis and Biot dimensionless groups as shown in Figure 4.

Figure 4. Fast pyrolysis regimes in Circulating Fluidized Bed, Static Fluidized Bed and Gas-Solid Vortex Reactor. The intrinsic kinetic regime corresponds to

and

and is confined to characteristic

lengths of less than 50 μm. For pyrolysis temperatures higher than 773 K and characteristics lengths of less than 0.4 mm, Figure 4 shows a region in which

and

. At these conditions the time

scale of heat transfer is smaller than that of chemical reactions and interfacial convection still controls

11

the process. The thermally thick regime is delimited by

and

; for these conditions the

time scale of heat transfer is smaller than that of chemical reactions and intraparticle conduction determines the observed reaction rate. The thermally thin regime is characterized by

and

; the time scale of chemical reactions is smaller than that of heat transfer and interfacial convection controls the process. Remark that the thermal wave regime with

and

falls

outside of the area in Figure 4 because of the relatively small particle diameters considered here. For particle diameters between 0.5 and 1.0 mm conventional FBs operate in the thermally thin regime. On the other hand, even for relatively small particle diameter, the GSVR operates in the thermally thick regime, i.e.,

and

. It could be argued that intraparticle conduction limitations

reduce the advantage from intensified interfacial convection in the GSVR [50]. However, the high interfacial convection in the GSVR rapidly increases the temperature at the surface of the particle thus creating a steep temperature gradient towards the center. This means that particles in the GSVR reach the pyrolysis temperature faster than in conventional FBs which has two important implications. First, the time during which biomass reacts at suboptimal temperatures is minimized leading to overall higher bio-oil yields. Second, this provides better control on the pyrolysis temperature which can be combined with detailed feedstock characterization and molecular-based kinetic models to steer the composition of the bio-oil [51, 52]. 3.2. Range of N2 and biomass mass flow rates For a given application, the reactor size and the operation conditions are mainly limited by carrier gas availability and thermal energy exchange toward or from the GSVR. In the fast pyrolysis case, mass and energy balances at steady state allow to estimate a preliminary range of biomass and N2 mass flow rates together with the corresponding operation temperature and thermal duty. The specified biomass feed composition resembles a typical woody material [53]: 40 wt. % cellulose, 23 wt. % hemicelluloses, 25 wt. % lignin, 8 wt. % moisture and 4 wt. % ashes. Biomass is assumed to be fully converted and the mass percentages on a dry basis of bio-oil, pyrolysis gas and solids are 74%, 9.0% and 17%, respectively. These mass percentages were obtained from previous CFD simulations of

12

biomass fast pyrolysis in a GSVR [20]. The enthalpies of reaction ΔHr for the production of bio-oil, char and pyrolysis gas are 255, -20 and -42 kJ kg-1, respectively [54-56]. Given the maximum N2 mass flow rate and the maximum N2 inlet temperature, the biomass mass flow rate is limited by the available thermal power to 1.4 g s-1. Calculations include 5% of heat losses to the surroundings and result in a thermal duty of 1.4 MJ kg-1 of dry biomass. This result is in agreement with the heat required for pyrolysis of biomass reported by Daugaard et al. [57] and Yang et al. [58] which varies from 1.0 to 1.6 MJ kg-1 on dry basis. Biomass mass flow rates for lower N2 mass flow rates at the maximum N2 inlet temperature are shown in Figure 5. A minimum biomass mass flow rate of 0.14 g s-1 allows to keep the SFA within 2.5 % as described in section 3.1. Previous cold flow experiments carried out in the first generation 54 cm diameter and 10 cm height gas-solid vortex unit (GSVU) have shown that gas injection velocities higher than 55 m s-1 are necessary to sustain a rotating fluidized bed [6]. Particles with densities of 950-1800 kg m-3 and diameters of 1-2 mm were used in those experiments. A N2 mass flow rate of 5 g s-1 allows to reach this gas injection velocity in the newly designed GSVR. This N2 mass flow rate is set as the lowest gas mass flow rate for design calculations and for the planning of fast pyrolysis experiments. Next, a maximum N2-to-biomass mass flow ratio of 36 is set to avoid excessive dilution of the reaction products. N2-to-biomass mass flow ratios higher than 36 reduce the bio-oil recovery. Those considerations result in the operation map shown in Figure 5. Section 3.3 presents a centrifugal-todrag force ratio analysis for the N2 mass flow rate range in Figure 5 and under expected reactive conditions.

13

Figure 5. Operation conditions in the Gas-Solid Vortex Reactor that fulfill the following requirements: 1) a 100 kg liquid N2 container suffices for 104 s runs, 2) the N2 inlet temperature does not exceed 930 K, 3) the solids feeding accuracy is within 2.5%, 4) the gas injection velocity is higher than 55 m s-1 and 5) the N2-to-biomass mass flow ratio does not exceed 36. The GSVR demonstration unit has been designed to operate with N2-to-biomass mass ratios of 7-36. The maximum N2 inlet temperature of 930 K set the lowest N2-to-biomass mass ratio. Implementing a HTL allows to decrease the N2-to-biomass mass ratio to 3-4, the typical ratio reported for FBs [59, 60] and ablative cyclone reactor pilot units [28]. A N2-to-biomass ratio of 2 has been reported for an spouted bed pilot plant [22]. On the other hand, the auger reactor and rotating cone reactor do not rely on carrier gas for bringing the solid heat carrier and the biomass into contact. For pilot units of those reactors the N2-to-biomass ratio varies from 0 to 0.25 [59]. When considering a commercial unit, it seems obvious that N2 would be replaced by the recycled non-condensable gases. Mullen et al. [61] found that partial or full replacement of N2 by fast pyrolysis non-condensable gases has a significant deoxygenation effect on the produced bio-oil. The GSVR targets valuable chemicals from biomass [62] so the influence of the composition of the carrier gas on bio-oil quality and yield will certainly be further investigated and exploited. 3.3 GSVR dimensions

14

The reactor dimensions and specific design details depend primarily on the drag force and centrifugal force, i.e. the two radially dependent and counteracting forces that act upon the rotating bed. The radially outward centrifugal force (

) is given in Equation 4 if the bed is assumed to rotate as a solid

body [63]. The radially inward drag force (

), is given by Equation 5 in which the Ergun equation is

used to describe the drag coefficient [63]. The Ergun equation is one of several correlations that can be used to describe the drag force. A semi-empirical drag correlation, specifically for the GSVU, is being developed [64]. If the drag force exceeds the centrifugal force, particles are entrained. Therefore, the centrifugal-to-drag force ratio gives a first indication of the feasibility of sustaining a rotating bed at the operation conditions determined in section 3.2. (4)

(5)

where represents bed void fraction, volumetric flow rate, diameter,

solids density,

number of reactor inlet-slots,

radial position,

gas viscosity,

azimuthal gas-solid slip factor, reactor length,

particle diameter and

These equations allow to specify the GSVR geometrical factors and operation conditions , void fraction

,

,

bed height,

gas

reactor

gas density. ,

, , and the gas and solid properties

based on the bed properties ,

,

,

. An average bed

of 0.6 is taken as reference based on the range observed in cold-flow experiments [8].

is set at 0.85 according to azimuthal gas-solid slip factors calculated via CFD simulations [63]. An average bed height The range of and

of 10 mm is specified in line with cold flow experiments reported in section 4.2.

corresponds to the feasible N2 mass flow rates shown in Figure 5. The effects of ,

on the centrifugal-to-drag force ratio are further assessed by varying their values within the

following ranges:

0.5-0.7,

0.75-0.95 and

The values for the carrier gas properties

and

7.5-15 mm. at 773 K and 130 kPa are accounted for via the

Chung correlation [65] and the ideal gas equation of state, respectively. The solids properties

15

and

require cautious considerations. First of all, CFD simulations of biomass fast pyrolysis in a GSVR indicate that, at steady state, char accounts for 97% of the mass of the bed [20]. The particle density of chars from different woody biomass varies between 300 and 430 kg m-3 [66]. Centrifugal-to-drag force ratios are evaluated for particle densities ranging from 250 to 500 kg m-3. The char morphology can be considered as a hollow and porous structure, caused by the quick release of volatiles during pyrolysis [67]. The char particle size is likely to be smaller than the original biomass due to shrinkage and break-up [68, 69]. When char particles are broken into smaller pieces, the density of the resultant particles tends to be larger than that of the original particle because of the partial elimination of its porous structure [49]. Centrifugal-to-drag force ratios are assessed for particle diameters ranging from 50 to 300 μm. For a given N2 mass flow rate, the momentum transferred to the bed is proportional to the square of the gas injection velocity, i.e. gas velocity at the outlet of the slots. The slots are thus manufactured as a single, easily replaceable piece. This ensures that the unit has the turndown required to handle a wide range of operation conditions for different gas-solid reactive processes. A reactor diameter mm and a reactor length-to-diameter ratio of 0.1875. i.e., reactor length

of 80

of 15 mm, are defined. The

latter is smaller than the maximum value of 0.5 recommended by Kochetov et al. to minimize nonuniformities in the axial direction [70]. For the geometry with 8 slots of 1 mm width oriented at 10° with respect to the tangent of the reactor chamber as shown in Figure 6, the ratio between the area occupied by the slots and the total circumferential area is 0.032. This value lies in the middle of the 0.025-0.038 range recommended by Kochetov et al. for relatively large Geldart B- and D-type particles [70]. The gas injection velocity for the N2 mass flow rates shown in Figure 5 varies from 55 to 140 m s-1.

16

Figure 6. Schematic of the Gas-Solid Vortex Reactor showing: a) top view of the gas feeding line, jacket and reactor chamber at a plane indicated by the cutting-plane line A-A, and b) side view of the jacket, reactor chamber and exhaust at a plane indicated by the cutting-plane line B-B. Biomass enters the GSVR next to the gas inlet slots through a circular conduit of 10 mm diameter. The biomass inlet is oriented at an 18° angle with respect to the horizontal plane due to space restrictions. Cross-sectional 2D and 3D views of the GSVR are shown in Figure S1 in the Supplementary Material. The bottom end wall is profiled according to the shape of the outlet nozzle in the top end wall of the reactor chamber. The diverging exhaust aims at minimizing the gas backflow and at preserving the swirling flow in the throat. The diverging exhaust wall profile follows an analytical solution for strongly swirling jets [71]. The design of the bottom end wall and exhaust profiles is based on the CFD simulations described in section 3.4. 17

The effects of bed height , bed void fraction

and azimuthal gas-solid slip factor

on the

centrifugal-to-drag force ratio are shown in Figure 7. Continuous and dashed lines correspond to pairs of particle size and particle density for which the centrifugal-to-drag force ratio equals unity at the inner edge of the bed as indicated in Figure 7a. The region at the right-hand side of the lines corresponds to pairs of particle size and particle density for which the centrifugal force exceeds the drag force at the inner edge of the bed as indicated in Figure 7e. The centrifugal-to-drag force ratio is calculated at the inner edge of the bed. With increasing gas flow rate, both centrifugal and drag force increase, but the former dominates. This is indicated by larger areas with centrifugal-to-drag force ratios higher than the unity in Figures 7d, 7e and 7f compared to those in Figures 7a, 7b, and 7c. Particle density exerts a smaller influence on the centrifugal-to-drag force ratio compared to the one exerted by particle size. As expected, the higher the bed height the lower the centrifugal force and the higher the risk of entrainment. Stable beds of more than 10 mm height and average particle diameters of less than 100 μm are unlikely. Overall, beds of 10 mm and average particle diameter of more than 200 μm are likely to be sustained. This particle size requires a reduction of 80% from a starting particle diameter of 1 mm. This considerable size reduction suggests that both shrinkage and break-up determine the average size of the particles retained in the bed. Bed void fraction and azimuthal gassolid slip factor exert a minor influence compared to that of bed height. For a constant bed height, higher ε and

values favor centrifugal force over drag force.

18

Figure 7. Influence of bed height , bed void fraction and azimuthal gas-solid slip factor

on the

centrifugal-to-drag force ratio. The base case is represented by continuous lines. Figures 7a, 7b and 7c relate to the lower limit of N2 mass flow rate, 5 g s-1. Figures 7d, 7e and 7f relate to the upper limit of N2 mass flow rates, 10 g s-1. Lines correspond to pairs of particle size and particle density for which the centrifugal-to-drag force ratio is unity at the inner edge of the bed as indicated in Figure 7a. The region at the right-hand side of the lines corresponds to pairs of particle size and particle density for which the centrifugal force exceeds the drag force at the inner edge of the bed as indicated in Figure 7e. Selective fine particles removal is one of the features of the GSVR to be harnessed in the case of fast pyrolysis. First, biomass particles undergo shrinking as a result of drying and the formation of an intermediate liquid [68, 72]. Then, primary fragmentation of the particles occurs as a consequence of thermal stresses and overpressure caused by volatiles emission. The remaining char is subjected to secondary fragmentation and attrition, exacerbated by its lower tensile strength. Figure 7 indicates that biomass particles smaller than 200 μm are at risk of being entrained. Results in Figure 7 apply for a geometry with 8 slots of 1 mm width oriented at 10° with respect to the tangent. The centrifugal-todrag force ratio can be modified by operating with a different slots ring, i.e. by changing the number 19

of slots, slot width or slot orientation with respect to the tangent. Because of its larger size, unreacted biomass is expected to occupy the outer edge of the bed. The largest char particles are likely to occupy the inner edge of the bed. The freeboard is expected to contain fines, either char or biomass, which are expected to be entrained as fresh biomass is continuously fed into the GSVR chamber. 3.4 Particle-free CFD simulations CFD simulations are used to assess the effect of the following GSVR design features: connection of the gas feeding line to the gas inlet jacket, shape of the gas inlet slots, bottom plate profile, shape of the diverging exhaust and connection of the diverging exhaust to the gas outlet line. Figure 8 shows the flow pattern in the final design via an azimuthal plane and cross-sectional plane colored by velocity magnitude and streamlines colored in grey. In contrast to previous designs [3, 5], there is only a single jacket feeding line (a), connected to a toroid with a rectangular cross-section surrounding the reaction chamber. Additionally, the connection of the gas feeding line to the jacket (b) features a slight decrease in the cross-sectional area, resulting in an increase in velocity of the fresh gas. This entrains the gas circulating in the jacket, similar to the working principle of an ejector. By applying a rotational motion to the gas in the jacket, the pressures at the entrance of the slots are inherently similar, given a sufficiently wide jacket. Similar pressures at the entrance of the slots implies an equal division of the gas flow over the available slots which aids bed stability. The calculated mass flow rates through each individual slot is within 3.5% relative standard deviation of the average. The edges of the slots (c) are rounded to minimize streamline curvature and the entrance of the slots is slightly enlarged to reduce the pressure drop in this high-gas-velocity region.

20

Figure 8. Flow pattern for gas-only flow in the Gas-Solid Vortex Reactor obtained via computational fluid dynamics simulations. Results correspond to an azimuthal plane and a cross-sectional plane colored by velocity magnitude and streamlines colored in grey. (a) Gas inlet, (b) connection of gas inlet to the jacket, (c) gas inlet slots, (d) profiled bottom end wall, (e) diverging exhaust, (f) gas outlet. Experiments and CFD simulations by Niyogi et al. have demonstrated the existence of a backflow region with negative axial velocity through the exhaust of the GSVR [9, 73]. The profiled bottom end wall (d) pushes the backflow region away from the reaction chamber through the exhaust to maximize the flow rate at the lowest possible pressure drop. The side wall of the diverging exhaust (e) is shaped according to an analytical solution for strongly swirling jets reported by Shtern and Hussain [71]. The angle of the diverging side wall with respect to its axis is a degree of freedom that was optimized via the CFD simulations in the range 30° - 45° to minimize the low velocity recirculation zone in the center of the diverging exhaust. A single, straight outlet (f) is connected tangentially to the diverging exhaust. The pressure profile in the GSVR is shown in Figure S2 in the Supplementary Material. The connection of the gas feeding line to the gas inlet jacket induces a pressure drop of 4.0 kPa. The slots

21

induce a pressure drop of 9.0 kPa. The relatively high pressure drop in the slots is intentional to have a damping effect in the presence of a solids bed in the GSVR [4]. Any upstream pressure fluctuations will be attenuated by the slots to enhance bed stability at steady-state operation. 4. Cold flow testing The purpose of this section is to assess the performance of the designed GSVR under particle-free and particulate flow conditions. Particle-free flow experiments focus on the capacity of the single main gas inlet to preserve pressure azimuthal symmetry in the jacket. Additionally, pressure measurements in the diverging exhaust indicate the strength of the low pressure region located at the center of this piece. Particulate flow experiments are carried out with the same biomass particles that will be used for high temperature experiments. Pressure profiles for the particle-free and particulate flow cases are then compared. Finally, the solids azimuthal velocity component is measured via PIV for three gas injection flow rates. This provides a first indication of the centrifugal force in the current GSVR and its dependency on the carrier gas injection velocity. 4.1 Particle-free flow Particle-free flow experiments at ambient temperature were carried out with air mass flow rates from 8.0 to 24 g s-1. Particle-free flow experiments were limited to air mass flow rates of up to 24 g s -1 because of the range of the absolute pressure sensor. The corresponding gas injection velocity varied from 50 to 110 m s-1 thus covering most of the gas injection velocity range of the operation map in Figure 5 (55-140 m s-1). Figure 9a shows that the pressure in the gas inlet jacket is evenly distributed in the azimuthal direction. Pressure distribution in the reactor chamber, both in the azimuthal and the radial direction, were assessed as well. Finally, pressure profiles at the diverging exhaust, at the center of the exhaust top and at the gas outlet were also measured.

22

Figure 9. Pressure profiles at different locations of the Gas-Solid Vortex Reactor for particle-free flow experiments at ambient temperature. a) Absolute pressure at the Jacket p1 position and pressure difference with respect to that at Jacket p1 at three different locations in the jacket, b) pressure difference with respect to that at Jacket p1 in the reactor chamber. Sensors were installed in two rows at different radii between 30 and 40 mm and separated 120° in the azimuthal direction. c) Pressure difference with respect to that at Jacket p1 at the diverging exhaust outer wall, exhaust top and gas outlet line, d) pressure sensor locations. Results correspond to the average of three repeated experiments, and error bars represent twice the standard deviation. Absolute pressure was measured at the jacket outer wall at 70° clockwise direction starting from the center of the gas feeding line. Pressure differences were measured at the jacket outer wall at azimuthal angles of 130°, 190° and 250° as schematized in Figure 9d. As shown in Figure 9a, the absolute

23

pressure continuously increased with increasing gas flow starting from 118 up to 159 kPa. The pressure at the 130° location slightly declined with a maximum decrement of 0.1 %. On the other hand, the pressures at the 190° and 250° locations continuously rose with a maximum increment of 0.3%. Figures 9a shows that a single gas feeding line combined with an ejector-like connection and the appropriate space in the jacket suffices for maintaining the azimuthal symmetry in the jacket. This confirms the CFD simulations described in section 3.4. Pressure differences with respect to that at Jacket p1 were measured at eight locations at the inner top surface of the reactor chamber. Figure 9b shows a slight difference in pressure at the inner wall attributed to the location of the sensors with respect to the slot outlet. On the other hand, for lower radii the pressure profiles at the two azimuthal locations are identical. The pressure drop through the slots increased continuously with increasing gas flow from 2.0 to 6.5 kPa. The pressure at the wall of the diverging exhaust decreased in the axial direction as shown in Figure 9c. The pressure measured at the outlet, which is tangential to the top end wall of the diverging exhaust, was identical to that at the middle of the diverging exhaust. On the other hand, Figure 9c shows a low pressure region at the center of the exhaust top end wall with an average pressure difference of 4 kPa with respect to that at the outlet. Particle-free flow CFD simulations showed that this low pressure region is indeed located in the diverging outlet around the axis of the geometry. However, the low pressure region is displaced upward through the exhaust away from the reactor chamber in the presence of the profiled bottom end wall. The overall particle-free flow pressure drop between the jacket and the outlet of the GSVR increased with increasing gas mass flow rate from 12 up to 51 kPa. 4.2 Particulate flow Particulate flow experiments at ambient temperature were carried out with air mass flow rates of 12-29 g s-1. The corresponding gas injection velocity varied from 75 to 140 m s-1. The current GSVR sustains a rotating bed of biomass particles in the full gas mass flow rate range mentioned above. It should be noted that particle-free flow experiments started from an air mass flow rate of 8 g s -1. Figure 10 shows that the presence of a rotating bed drastically changes the pressure profiles in the reactor chamber and

24

in the exhaust. This explains why a wider gas mass flow rate was covered for particulate flow experiments while using the same type of pressure sensors.

Figure 10. Pressure profiles at different locations of the Gas-Solid Vortex Reactor for particulate flow experiments. a) Absolute pressure at the Jacket p1 position and pressure difference with respect to that at Jacket p1 at one position in the jacket, inner wall of the reactor chamber and inner edge of the rotating bed. b) Pressure differences with respect to that at Jacket p1 at the diverging exhaust outer wall, exhaust top and gas outlet line. Results correspond to the average of three repeated experiments, and error bars represent twice the standard deviation. Absolute pressure continuously increased with increasing gas mass flow rate starting from 112 up to 139 kPa. Pressure was evenly distributed in the gas inlet jacket as shown in Figure 10a, similar to the particle-free flow case shown in Figure 9a. The pressure drop through the slots increased continuously with increasing gas flow from 4.9 to 15 kPa. A similar trend was observed for the pressure drop through the bed which increased continuously with increasing gas flow from 0.50 to 1.7 kPa. In contrast to particle-free flow experiments, there was no substantial pressure drop in the axial direction of the diverging exhaust and the low pressure region at the center of the exhaust top was not detected as shown in Figure 9b. Besides, the overall pressure drop between the jacket and the outlet increased with increasing gas flow from 6.6 up to 26 kPa. This is a four-fold decrease in pressure drop for equivalent gas mass flow rates compared to the case of particle-free flow. This allowed to extend the air mass flow rate range to 29 g s-1 compared to the maximum value of 24 g s-1 for particle-free flow. 25

The air mass flow was limited in this case because of the maximum capacity of the flow meter. A similar result have been obtained in a 54 cm diameter and 10 cm height GSVU in which a five-fold decrease in pressure drop between particle-free and particulate flow was measured [9]. The advantages of improved interfacial energy, mass and momentum transfer, together with bed uniformity and enhanced temperature control in the GSVR come at the cost of a significant pressure drop. Figure 10 shows that the pressure drop through the slots accounts for 60-73% of the total pressure drop for particulate flow. The contribution of the slots to the total pressure drop decreases with increasing air mass flow rate. Note that there is still room for decreasing the total pressure drop by increasing the exhaust throat diameter and by further optimizing the geometry of the slots. Figure 11 shows the solids azimuthal velocity for three different gas injection velocities. The average solids azimuthal velocity for 20 circular lines corresponding to constant radius between 25 mm and 38 mm has been calculated, and used to construct the radial profiles of solids azimuthal velocity. The relative standard deviation for the solids velocity calculated at each radii was within 20%. This variation in the solids azimuthal velocity cannot be visually observed. The maximum in the solids azimuthal velocity downstream the circumferential wall have also been observed in previous experiments in a GSVU of 54 cm diameter and 10 cm length [7]. Particulate flow experiments are representative for rotating beds without particles entrainment over the edge of the bed.

26

Figure 11. Solids azimuthal velocity across the rotating bed of pinewood with an average particle density of 500 kg m-3 and maximum dimension of 1.5 mm. a) Radial distribution of solids azimuthal velocity for three air injection velocities. The PIV section analyzed is the bottom end wall section from 190° to 242°, radial positions from r = 25 mm to r = 38 mm. There was no optical access to the solids bed over a radius of 38 mm, due to the metallic edge that holds the transparent plate. Results correspond to the average of three repeated experiments, each of them consisting of 100 PIV pairs, and error bars represent twice the standard deviation. Solids azimuthal velocity increases with increasing gas injection velocity as the momentum input increases as well. The biomass loading varied from 8 to 10 g for the same type of biomass particles. The instantaneous bed images as shown in Figure 11b and on the attached videos (see Supplementary Material), indicate that a compact bed can be sustained. Results in Figure 10b corresponds to a gas injection velocity of 95 m s-1. Average bed heights varied from 9 mm to 11 mm (std < 3%). Higher gas injection velocities induce a higher gas-to-solids momentum transfer, allowing higher solids azimuthal velocity, as can be observed in Figure 10a. This results in an increase of the solids centrifugal force and consequently in higher solids mass capacities. Figure 11 shows that the current GSVR sustains rotating beds with solids azimuthal velocity similar to the ones observed in a GSVU of 6.8 times larger diameter [7]. By assuming solid body rotation, the centrifugal force per unit volume experienced by 27

the bed in the current GSVR is 6.8 times higher as well. On the other hand, the superficial gas velocities are in the same range 2.5-5.0 m s-1 which indicates that there are no substantial differences in drag force. 5. Conclusions This work presents the design and cold flow testing of a GSVR for biomass fast pyrolysis. Gas-solid slip velocities in the GSVR are much higher than those in conventional FBs which leads to intensified interfacial transfer of heat, mass and momentum. The reactor chamber has a dimeter of 80 mm and a height of 15 mm, which allows to work with carrier gas mass flow rates between 5-10 g s-1 and biomass mass flow rates between 0.14-1.4 g s-1 at biomass fast pyrolysis conditions. Rotating beds of up to 10 mm height with biomass particles differing in shape and size are sustained. The GSVR operates in the thermally thick regime, which leads to an improved particle temperature control and to the minimization of intraparticle volumes with suboptimal temperatures. CFD simulations and experiments showed that the specific carrier gas inlet design guarantees that pressure is evenly distributed at the entrances of the slots. Solids retention and operation stability is further improved by including a profiled bottom end wall and a diverging exhaust. Cold flow experiments with pinewood particles of 1.5 mm maximum dimension confirmed that the GSVR can sustain a rotating bed within a wide range of carrier gas mass flow rates. Particulate flow experiments also showed a pressure drop through the slot ten times higher than that through the bed. Radial profiles of solids azimuthal velocity measured with 2D PIV demonstrated that the GSVR sustains rotating beds with average solids azimuthal velocity similar to the ones observed in a GSVU of 6.8 times larger diameter. On the other hand, superficial gas velocities are in the same range which indicates that there are no substantial differences in drag force. As a result, the designed GSVR achieves a sufficiently high centrifugal-todrag force ratio sustaining a rotating fluidized bed within a broad range of operation conditions and shows great potential for biomass fast pyrolysis and related processes. Supporting Material

28

Pyrolysis gas and bio-oil molecular composition. Schematics showing 2D and 3D views of the GasSolid Vortex Reactor. Particle-free flow visualization of the pressure distribution in the Gas-Solid Vortex Reactor. Acknowledgements The research leading to these results has received funding from the European Research Council under the European Union’s Seventh Framework Programme FP7/2007-2013/ERC grant agreement n° 290793 and the‘Long Term Structural Methusalem Funding by the Flemish Government’. This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 664876. The SBO proposal ‘‘Bioleum” supported by the Institute for Promotion of Innovation through Science and Technology in Flanders (IWT) is acknowledged. The computational work was carried out using the STEVIN Supercomputer Infrastructure at Ghent University, funded by Ghent University, the Flemish Supercomputer Center (VSC), the Research Foundation - Flanders (FWO) and the Flemish Government – department EWI. Pieter Reyniers acknowledges financial support from a doctoral fellowship from the Research Foundation - Flanders (FWO). Dr. Vladimir Shtern is acknowledged for the scientific discussions on swirling jets and the GSVR diverging exhaust.

Nomenclature

CFB CFD

FB FCC

Biot number – (hpL)/λs Circulating fluidized beds Computational fluid dynamics Biomass heat capacity (J kg-1 K-1) Particle diameter (mm) First Damköhler number – (ρsCpsL2k)/λs Second Damköhler number – (ρsCpsLk)/hp Exhaust diameter (mm) Jacket diameter (mm) Reactor diameter (mm) fluidized bed Fluid catalytic cracking Centrifugal force per unit volume (N m-3) Drag force per unit volume (N m-3) 29

GSVR GSVU

HCL

NREL PIV

QUICK

RANS RSM SFA SFB

VOC

Gas-solid vortex reactor Gas-solid vortex unit Bed height (mm) Convective heat transfer coefficient (W m-2 K-1) Heat carrier loop Slot thickness (mm) Pyrolysis reaction rate coefficient (s-1) Reactor length (mm) Characteristic length scale (m) Inlet gas mass flow rate (g s-1) Outlet gas mass flow rate (g s-1) Inlet solids mass flow rate (g s-1) Outlet solids mass flow rate (g s-1) Number of reactor inlet-slots (-) National Renewable Energy Laboratory particle image velocimetry Pyrolysis number I – λs/(ρsCpsL2k) Pyrolysis number II – hp/(ρsCpsLk) Quadratic Upstream Interpolation for Convective Kinematics Gas volumetric flow rate (m3 s-1) Radial position (mm) Reynolds-averaged Navier-Stokes Reynolds Stress Model Solids feeding accuracy Static fluidized beds Biomass inlet temperature (K) Solid phase pyrolysis temperature (K) Gas inlet temperature (K) Solid phase outlet temperature (K) Volatile organic compounds Bed void fraction (-) Enthalpy of reaction (kJ kg-1) Gas density (kg m-3) Solids density (kg m-3) Gas injection angle (°) Biomass thermal conductivity (W m-1 K-1) Gas viscosity (kg m-1 s-1) Azimuthal gas-solid slip factor (-)

References [1] A.I. Stankiewicz, J.A. Moulijn, Process Intensification: Transforming Chemical Engineering, Chem. Eng. Prog., 96 (2000) 22-34. [2] Z. Wei, A Review of Techniques for the Process Intensification of Fluidized Bed Reactors, Chi. J. Chem. Eng., 17 (2009) 688-702. [3] J. De Wilde, A. de Broqueville, Rotating fluidized beds in a static geometry: Experimental proof of concept, AIChE J., 53 (2007) 793-810. [4] A. Dutta, R.P. Ekatpure, G.J. Heynderickx, A. de Broqueville, G.B. Marin, Rotating fluidized bed with a static geometry: Guidelines for design and operating conditions, Chem. Eng. Sci., 65 (2010) 1678-1693. 30

[5] R.P. Ekatpure, V.U. Suryawanshi, G.J. Heynderickx, A. de Broqueville, G.B. Marin, Experimental investigation of a gas–solid rotating bed reactor with static geometry, Chem. Eng. Process. Process Intensif., 50 (2011) 77-84. [6] J.Z. Kovacevic, M.N. Pantzali, G.J. Heynderickx, G.B. Marin, Bed stability and maximum solids capacity in a Gas–Solid Vortex Reactor: Experimental study, Chem. Eng. Sci., 106 (2014) 293-303. [7] J.Z. Kovacevic, M.N. Pantzali, K. Niyogi, N.G. Deen, G.J. Heynderickx, G.B. Marin, Solids velocity fields in a cold-flow Gas–Solid Vortex Reactor, Chem. Eng. Sci., 123 (2015) 220-230. [8] M.N. Pantzali, J.Z. Kovacevic, G.J. Heynderickx, G.B. Marin, V.N. Shtern, Radial pressure profiles in a cold-flow gas-solid vortex reactor, AIChE J., 61 (2015) 4114-4125. [9] K. Niyogi, M. Torregrosa, M.N. Pantzali, G.J. Heynderickx, G.B. Marin, Experimentally validated numerical study of gas-solid vortex unit hydrodynamics, Powder Technol., 305 (2017) 794-808. [10] R.H. Venderbosch, W. Prins, Fast pyrolysis technology development, Biofuels, Bioprod. Biorefin., 4 (2010) 178-208. [11] L. Negahdar, A. Gonzalez-Quiroga, D. Otyuskaya, H.E. Toraman, L. Liu, J.T.B.H. Jastrzebski, K.M. Van Geem, G.B. Marin, J.W. Thybaut, B.M. Weckhuysen, Characterization and Comparison of Fast Pyrolysis Bio-oils from Pinewood, Rapeseed Cake, and Wheat Straw Using 13C NMR and Comprehensive GC× GC, ACS Sustainable Chem. Eng., 4 (2016) 4974-4949-4985. [12] M.R. Djokic, T. Dijkmans, G. Yildiz, W. Prins, K.M. Van Geem, Quantitative analysis of crude and stabilized bio-oils by comprehensive two-dimensional gas-chromatography, J. Chromatogr. A, 1257 (2012) 131-140. [13] J. De Wilde, Gas–solid fluidized beds in vortex chambers, Chem. Eng. Process. Process Intensif., 85 (2014) 256-290. [14] W. Rosales Trujillo, J. De Wilde, Fluid catalytic cracking in a rotating fluidized bed in a static geometry: a CFD analysis accounting for the distribution of the catalyst coke content, Powder Technol., 221 (2012) 36-46. [15] P. Eliaers, A. de Broqueville, A. Poortinga, T. van Hengstumb, J. De Wilde, High-G, lowtemperature coating of cohesive particles in a vortex chamber, Powder Technol., 258 (2014) 242-251. [16] R.W. Ashcraft, J.Z. Kovasevic, G.J. Heynderickx, G.B. Marin, Assessment of a Gas−Solid Vortex Reactor for SO2/NOx Adsorption from Flue Gas, Ind. Eng. Chem. Res., 52 (2013) 861-875. [17] L.M. Kochetov, B.S. Sazhin, E.A. Karlik, Hydrodynamics and heat exchange in vortex drying chambers, Chem. Petrol. Eng., 5 (1969) 688-690. [18] N. Staudt, A. de Broqueville, W. Rosales Trujillo, J. De Wilde, Low-Temperature Pyrolysis and Gasification of Biomass: Numerical Evaluation of the Process Intensification Potential of Rotatingand Circulating Rotating Fluidized Beds in a Static Fluidization Chamber, Int. J. Chem. React. Eng., 9 (2011) 1-31. [19] E.P. Volchkov, N.A. Dvornikov, V.V. Lukashov, R.K. Abdrakhmanov, Investigation of the flow in the vortex chamber with centrifugal fluidizing bed with and without combustion, Thermophys. Aeromec., 20 (2013) 663-668. [20] R.W. Ashcraft, G.J. Heynderickx, G.B. Marin, Modeling fast biomass pyrolysis in a gas–solid vortex reactor, Chem. Eng. J., 207 (2012) 195-208. [21] A. Gonzalez-Quiroga, H.-H. Carstensen, K.M. Van Geem, G.B. Marin, Design of a gas-solid vortex reactor demonstration unit for the fast pyrolysis of lignocellulosic biomass, in: 21st International Symposium on Analytical and Applied Pyrolysis (PYRO 2016), Nancy, 2016, pp. 181181. [22] A.R. Fernandez-Akarregi, J. Makibar, G. Lopez, M. Amutio, M. Olazar, Design and operation of a conical spouted bed reactor pilot plant (25 kg/h) for biomass fast pyrolysis, Fuel Process. Technol., 112 (2013) 48-56. [23] Y. Solantausta, A. Oasmaa, K. Sipila, C. Lindfors, J. Lehto, J. Autio, P. Jokela, J. Alin, J. Heiskanen, Bio-oil production from biomass: steps toward demonstration, Energy Fuels, 26 (2012) 233-240. [24] A. Oasmaa, B. van de Beld, P. Saari, D.C. Elliott, Y. Solantausta, Norms, Standards, and Legislation for Fast Pyrolysis Bio-oils from Lignocellulosic Biomass, Energy Fuels, 29 (2015) 24712484. [25] B.M. Wagenaar, W. Prins, W.P.M. Van Swaaij, Pyrolysis of biomass in the rotating cone reactor: modelling and experimental justification, Chem. Eng. Sci., 49 (1994) 5109-5126. 31

[26] R. Venderbosch, W. Prins, Fast pyrolysis of biomass for energy and chemicals: technologies at various scales, in: J. Hamsen, J.B. Powell (Eds.) Sustainable Development in the Process Industries, Wiley, New Jersey, 2010, pp. 109-155. [27] . Pfitzer, N. ahmen, N. Tr ger, F. Weirich, J.r. Sauer, A. G nther, M. M ller-Hagedorn, Fast Pyrolysis of Wheat Straw in the Bioliq Pilot Plant, Energy Fuels, 30 (2016) 8047-8054. [28] H. Wiinikka, P. arlsson, A. Johansson, M. Gullberg, . lip , M. Lundgren, L. Sandstr m, Fast Pyrolysis of Stem Wood in a Pilot-Scale Cyclone Reactor, Energy Fuels, 29 (2015) 3158-3167. [29] R.S. Miller, J. Bellan, Numerical Simulation of Vortex Pyrolysis Reactors for Condensable Tar Production from Biomass, Energy Fuels, 12 (1998) 25-40. [30] J. Lédé, Biomass Fast Pyrolysis Reactors: A Review of a Few Scientific Challenges and of Related Recommended Research Topics, Oil Gas Sci. Technol. – Rev. IFP Energies nouvelles, 68 (2013) 801-814. [31] T. Kan, V. Strezov, T.J. Evans, Lignocellulosic biomass pyrolysis: A review of product properties and effects of pyrolysis parameters, Renewable Sustainable Energy Rev., 57 (2016) 1126-1140. [32] J. Dai, H. Cui, J.R. Grace, Biomass feeding for thermochemical reactors, Prog. Energy Combust. Sci., 38 (2012) 716-736. [33] A.S. Pollard, M.R. Rover, R.C. Brown, Characterization of bio-oil recovered as stage fractions with unique chemical and physical properties, J. Anal. Appl. Pyrolysis, 93 (2012) 129-138. [34] D. Edwards, Scaling Up Bioenergy Technologies, Chem. Eng. Prog., 111 (2015) 58-61. [35] E. Henrich, N. Dahmen, F. Weirich, R. Reimert, C. Kornmayer, Fast pyrolysis of lignocellulosics in a twin screw mixer reactor, Fuel Process. Technol., 143 (2016) 151-161. [36] D. Mohan, C.U. Pittman, P.H. Steele, Pyrolysis of Wood/Biomass for Bio-oil: A Critical Review, Energy Fuels, 20 (2006) 848-889. [37] A.V. Bridgwater, Review of fast pyrolysis of biomass and product upgrading, Biomass Bioenergy, 38 (2012) 68-94. [38] G. Towler, R.K. Sinnott, Chemical Engineering Design: Principles, Practice and Economics of Plant and Process Design, Elsevier, 2012. [39] R.J. Bedmutha, L. Ferrante, C. Briens, F. Berrutia, I. Inculet, Single and two-stage electrostatic demisters for biomass pyrolysis application, Chem. Eng. Process. Process Intensif., 48 (2009) 11121120. [40] S. Jones, P. Meyer, L. Snowden-Swan, A. Padmaperuma, E. Tan, A. Dutta, J. Jacobson, K. Cafferty, Process Design and Economics for the Conversion of Lignocellulosic Biomass to Hydrocarbon Fuels Fast Pyrolysis and Hydrotreating Bio-oil Pathway, in, NREL TP-5100-61178, 2013. [41] USEPA, Air Pollution Control Cost Manual, EPA/452/B-02-001, 2002. [42] D.L. Pyle, C.A. Zaror, Heat transfer and kinetics in the low temperature pyrolysis of solids, Chem. Eng. Sci., 39 (1984) 147-158. [43] J. Chang, G. Wang, J. Gao, K. Zhang, H. Chen, Y. Yang, CFD modeling of particle–particle heat transfer in dense gas-solid fluidized beds of binary mixture, Powder Technology, 217 (2012) 50-60. [44] Z. Shu, J. Wang, Q. Zhou, C. Fan, S. Li, Evaluation of multifluid model for heat transfer behavior of binary gas–solid flow in a downer reactor, Powder Technology, 281 (2015) 34-48. [45] D.J. Gunn, Transfer of heat or mass to particles in fixed and fluidised beds, Int. J. Heat Mass Transfer, 21 (1978) 467-476. [46] H. Lu, E. Ip, J. Scott, P. Foster, M. Vickers, L.L. Baxter, Effects of particle shape and size on devolatilization of biomass particle, Fuel, 89 (2010) 1156-1168. [47] H. Tavassoli, E.A.J.F. Peters, J.A.M. Kuipers, Direct numerical simulation of fluid–particle heat transfer in fixed random arrays of non-spherical particles, Chem. Eng. Sci., 129 (2015) 42-48. [48] Q. Xue, T.J. Heindel, R.O. Fox, A CFD model for biomass fast pyrolysis in fluidized-bed reactors, Chem. Eng. Sci., 66 (2011) 2440-2452. [49] W.C. Yang, Handbook of fluidization and fluid-particle systems, CRC press, 2003. [50] J.R. Pati, S. Dutta, P. Eliaers, P. Mahanta, P.K. Chatterjee, J. De Wilde, Experimental study of paddy drying in a vortex chamber, Drying Technol., 34 (2016) 1073-1084. [51] R. Van de Vijver, B.R. Devocht, K.M. Van Geem, J.W. Thybaut, G.B. Marin, Challenges and opportunities for molecule-based management of chemical processes, Curr. Opin. Chem. Eng., 13 (2016) 142-149. 32

[52] X. Zhou, M.W. Nolte, H.B. Mayes, B.H. Shanks, L.J. Broadbelt, Experimental and Mechanistic Modeling of Fast Pyrolysis of Neat Glucose-Based Carbohydrates. 1. Experiments and Development of a Detailed Mechanistic Model, Ind. Eng. Chem. Res., 53 (2014) 13274-13289. [53] S.V. Vassilev, D. Baxter, L.K. Andersen, C.G. Vassileva, T.J. Morgan, An overview of the organic and inorganic phase composition of biomass, Fuel, 94 (2012) 1-33. [54] F. Shafizadeh, A. Bradbury, Thermal degradation of cellulose in air and nitrogen at low temperatures, J. Appl. Polym. Sci., 23 (1979) 1431-1442. [55] S.M. Ward, J. Braslaw, Experimental weight loss kinetics of wood pyrolysis under vacuum, Combust. Flame, 61 (1985) 261-269. [56] F.J. Kilzer, A. Broido, Speculations on the nature of cellulose pyrolysis, Pyrodynamics, 2 (1965) 151-163. [57] D.E. Daugaard, R.C. Brown, Enthalpy for Pyrolysis for Several Types of Biomass, Energy Fuels, 17 (2003) 934-939. [58] H. ang, S. Kudo, H. Kuo, K. Norinaga, A. Mori, O. Ma ek, J. Hayashi, Estimation of Enthalpy of Bio-Oil Vapor and Heat Required for Pyrolysis of Biomass, Energy Fuels, 27 (2013) 2675-2686. [59] J. Joubert, M. Carrier, N. Dahmen, R. Stahl, J.H. Knoetze, Inherent process variations between fast pyrolysis technologies: A case study on Eucalyptus grandis, Fuel Process. Technol., 131 (2015) 389-395. [60] J.-Y. Jeong, U.-D. Lee, W.-S. Chang, S.-H. Jeong, Production of bio-oil rich in acetic acid and phenol from fast pyrolysis of palm residues using a fluidized bed reactor: Influence of activated carbons, Bioresource Technol., 219 (2016) 357-364. [61] C.A. Mullen, A.A. Boateng, N.M. Goldberg, Production of Deoxygenated Biomass Fast Pyrolysis Oils via Product Gas Recycling, Energy Fuels, 27 (2013) 3867-3874. [62] A. Gonzalez-Quiroga, K.M. Van Geem, G.B. Marin, Towards first-principles based kinetic modeling of biomass fast pyrolysis, Biomass Conv. Bioref., (2017) 1-13, doi: 10.1007/s13399-1301710251-13390. [63] A. De Broqueville, J. De Wilde, Numerical investigation of gas-solid heat transfer in rotating fluidized beds in a static geometry, Chem. Eng. Sci., 64 (2009) 1232-1248. [64] M. Friedle, K. Niyogi, M.M. Torregrosa, G.J. Heynderickx, G.B. Marin, Semi-empirical drag correlation for a Gas Solid Vortex Reactor starting from the radial momentum balances, in: AIChE Annual Meeting, San Francisco, 2016. [65] T.H. Chung, M. Ajlan, L.L. Lee, K.E. Starling, Generalized multiparameter correlation for nonpolar and polar fluid transport properties, Ind. Eng. Chem. Res., 27 (1988) 671-679. [66] A. Downie, P. Munroe, A. Crosky, Characteristics of biochar - physical and structural properties, in: J. Lehman, S. Josep (Eds.) Biochar for environmental management: Science and technology, Earthscan, London, 2009, pp. 13-29. [67] K.H. Kim, I.Y. Eom, S.M. Lee, D. Choi, H. Yeo, I. Choi, J.W. Choi, Investigation of physicochemical properties of biooils produced from yellow poplar wood (Liriodendron tulipifera) at various temperatures and residence times, J. Anal. Appl. Pyrolysis, 92 (2011) 2-9. [68] A.D. Paulsen, B.R. Hough, C.L. Williams, A.R. Teixeira, D.T. Schwartz, J. Pfaendtner, P.J. Dauenhauer, Fast Pyrolysis of Wood for Biofuels: Spatiotemporally Resolved Diffuse Reflectance In situ Spectroscopy of Particles, ChemSusChem, 7 (2014) 765-776. [69] K.H. Kim, J. Kim, T. Cho, J.W. Choi, Influence of pyrolysis temperature on physicochemical properties of biochar obtained from the fast pyrolysis of pitch pine (Pinus rigida), Bioresour. Technol., 118 (2012) 158-162. [70] L.M. Kochetov, B.S. sazhin, E.A. karlik, Experimental determination of the optimal ratios of structural dimensions in the whirl chamber fro drying granular materials, Chem. Petrol. Eng., 5 (1969) 106-108. [71] V. Shtern, F. Hussain, Hysteresis in a swirling jet as a model tornado, Phys. Fluids A, 5 (1993) 2183-2195. [72] M.S. Mettler, D.G. Vlachos, P.J. Dauenhauer, Top ten fundamental challenges of biomass pyrolysis for biofuels, Energy Environ. Sci, 5 (2012) 7797-7809. [73] K. Niyogi, M.M. Torregrosa, M.N. Pantzali, G.J. Heynderickx, G.B. Marin, V.N. Shtern, On near‐wall jets in a disc‐like gas vortex unit, AIChE J. doi: 10.1002/aic.15533, (2016).

33

Highlights 

Design of a Gas-Solid Vortex Reactor or GSVR for biomass fast pyrolysis



New features: single main gas inlet, profiled bottom end wall and diverging exhaust



Sufficiently high centrifugal-to-drag force ratio for sustaining a rotating bed



Stable rotating fluidized bed with average solids azimuthal velocities of 6-7 m s-1

34