The Structure of Supercritical Fluid Free-Jet Expansions Imane Khalil and David R. Miller Dept. of Mechanical and Aerospace Engineering, University of California at San Diego, La Jolla, CA 92093 DOI 10.1002/aic.10285 Published online in Wiley InterScience (www.interscience.wiley.com). Experimental and numerical studies are reported on the structure of supersonic free-jet expansions of supercritical CO2 impacting on a flat plate. Numerical calculations for the axisymmetric, two-dimensional (2-D) expansion use a time-dependent finite difference method known as the two-step Lax-Wendroff technique, incorporating the Redlich-Kwong equation of state to model CO2. The numerical results are compared with experimental optical shadowgraph measurements of the jet and shock wave structure, impact pressure and temperature measurements along the plate, and a thermocouple probe of the expan- sion. Approximations based on ideal gases and quasi-1-D flow analysis, often used by researchers, are found to be useful for these supercritical fluid flows. © 2004 American Institute of Chemical Engineers AIChE J, 50: 2697–2704, 2004 Introduction sonic conditions, Mach number equal to one, near the exit. The fluid then expands in a highly compressible, 2-D, supersonic, There has been considerable interest in the supersonic free-jet expansion, terminated by shock waves as the fluid pres- free-jet expansion of supercritical fluids (SCF) because of sure adjusts to ambient conditions. These free-jet expansions have the role this rapid expansion plays in the decompression of been well studied for ideal gases and utilized by molecular beam high-pressure supercritical fluid mixtures. The rapid expan- researchers for more than 40 years.4 The purpose of this research sion of supercritical solutions (RESS) has been studied for is to theoretically and experimentally examine the structure of the nearly 20 years since the early articles of Smith and cowork- free-jet expansion for a SCF. 1 ers to fabricate nanoscale particles, grow thin films, and as An excellent overview of the RESS process, including the an interface to instrumentation, such as mass spectrometers fluid mechanics of the subsonic and supersonic flows, is pro- and gas chromatographs. Several reviews2,3 are available vided by Weber and Thies5 and earlier by Debenedetti and which describe the properties and uses of SCF. In our colleagues.6 These authors have borrowed results from the laboratory we are interested in using the properties of su- ideal gas free-jet literature and utilized quasi-1-D (QOD) anal- percritical fluids to dissolve nonvolatile or temperature sen- ysis of the free-jet as an approximation to understand the RESS sitive materials into a SCF and then to ultimately extract molec- expansion for supercritical fluids. In this article we present ular beams from the free-jet expansions, which will permit us to numerical results for the axisymmetric, 2-D, free-jet (ASFJ) examine the structure and chemical physics of small clusters. The expansion with a Redlich-Kwong equation of state, which high-pressure SCF mixture is decompressed typically through a incorporates the repulsive potential excluded volume and the narrow orifice or a long capillary tube. As the SCF flows through attractive potential real gas effects, and is a first approximation the orifice or tube, it accelerates in subsonic flow and reaches for many SCFs. We also present experimental data for super- critical CO2 expansions, which compares reasonably well with our calculations for the jet structure. The expansion is directed Correspondence concerning this article should be addressed to D. R. Miller at [email protected]. at a flat plate, to simulate the use of RESS to grow thin films. Current address of I. Khalil: Sandia National Laboratories, Albuquerque, NM; We find that the QOD analysis and the ideal gas approximation e-mail: [email protected]. do in fact provide reasonable first approximations to the flow © 2004 American Institute of Chemical Engineers properties. AIChE Journal November 2004 Vol. 50, No. 11 2697 We have primarily studied the free-jet expansion of pure ␳ CO2 originating from small orifices rather than capillary ␳u E ϭ tubes in order to reduce the effects of viscosity, heat trans- Έ ␳v Έ fer, and clustering or condensation in the subsonic flow. The ␳͓e ϩ 1/2͑u2 ϩ v2͔͒ subsonic expansion from stagnation to sonic conditions in the orifice flow is essentially inviscid and adiabatic, hence, isentropic, which provides a well defined inlet boundary ␳u condition for our free-jet calculations and for our experi- ␳u2 ϩ P F ϭ ments. In the later sections we first briefly discuss the Έ ␳uv Έ equations and our numerical method, and then the experi- u͕␳͓e ϩ 1/2͑u2 ϩ v2͔͒ ϩ P͖ mental methods and results. Additional details are available elsewhere.7 ␳v ␳uv G ϭ Έ ␳v2 ϩ P Έ Theory and Numerical Calculations v͕␳͓e ϩ 1/2͑u2 ϩ v2͔͒ ϩ P͖ For ideal gases, there have been many rigorous calcula- 4,8 tions of axisymmetric free-jet supersonic expansions. ␳v Although the method of characteristics is regarded as the ␳uv most accurate for the isentropic supersonic free-jet expan- H ϭ 2 Έ ␳v Έ sion, the time-dependent methods are useful to correctly v͕␳͓e ϩ 1/2͑u2 ϩ v2͔͒ ϩ P͖ capture shock wave structure and to include kinetic effects,9 and they can be extended to the viscous subsonic flow. Many texts provide recipes and summarize the advantages and disadvantages of various numerical methods applicable to In addition, two equations of state are required, e(␳,T) and the free-jet supersonic expansion.10 One of the difficulties of P(␳,T), to complete the set of six equations and six unknowns, incorporating real gas effects into the supersonic compress- with velocity components u and v in the axial and radial ible flow calculation is the need to calculate the speed of directions, respectively, pressure P, temperature T, density ␳, sound at each grid point and time step, especially important and the internal energy per unit mass e. We also use enthalpy in methods which incorporate compression and rarefaction h(␳,T) ϭ e ϩ P/␳, to compute e(␳,T), and entropy s(␳,T), for waves, such as the flux splitting Godunov type methods. isentropic calculations. Except near solid boundaries and in the Often for hypersonic nozzles, wherein flows originate from jet boundary shear layers, the neglect of viscosity and heat high temperature and pressure, it is sufficient to consider conduction is a good approximation. For the SCFs and condi- only the excluded volume repulsive potential correction to tions considered here, the more serious approximation is that ideal gas behavior.11 However, since the SCF flows of we assume a homogeneous fluid, neglecting clustering and interest here originate from high pressure and sufficiently condensation effects. The numerical method can incorporate low temperatures, it is necessary to consider both the ex- such kinetics, however, we defer such analysis to a future cluded volume and the attractive potential corrections to publication. ideal gas behavior. We have chosen the finite difference, We have studied several equations of state in our calcula- two-step Lax-Wendroff method, because it is suitable for tions, including Redlich-Kwong and Peng-Robinson cubic the hyperbolic partial differential equations of the free-jet, it equations,16 as well as Huang et al.’s 27-parameter equation,17 has been utilized successfully for ideal gas free-jet expan- and we previously showed that they provide similar results for 7,18 sions, the real fluid equations of state are easily incorpo- CO2. Although the temperature-dependent parameters make rated, and the speed of sound enters explicitly only in the the Peng-Robinson and 27-parameter equations of state more stability criteria for the time increment.12 We begin by rigorous, especially at higher pressures, the axisymmetric re- sults presented here primarily use the simple Redlich-Kwong studying free-jet expansions for pure supercritical CO2 be- cause it is a well studied fluid in both the molecular beam cubic equation of state field, under ideal gas conditions, and in the SCF field as an important supercritical solvent. Furthermore, CO2 has been RT a characterized by several equations of state for which the ϭ Ϫ P Ϫ 1/ 2 ͑ ϩ ͒ (2) interaction parameters are well determined in the litera- v b T v v b ture.13,14 The time-dependent partial differential equations to be 2 2.5 R Tc RTc solved for the inviscid, adiabatic, axisymmetric free-jet super- a ϭ 0.42784 and b ϭ 0.08664 P P sonic expansion (ASFJ) are well established8,15 and are given c c in conservation form in Eqs. 1 The cubic equations of state are readily adaptable to many mix- tures of solutes with solvents, they have been well studied by chemical engineers, and the mixture interaction parameters are Ѩ Ѩ Ѩ E F G H available.13,14 Calculations of e, h, and s are taken from standard ϩ ϩ ϩ ϭ 0 (1) Ѩt Ѩ x Ѩr r texts.16 As an example, the Redlich-Kwong relation for h becomes 2698 November 2004 Vol. 50, No. 11 AIChE Journal the classically excited translation and rotation degrees of free- dom, which provides a constant heat capacity Cp ϭ 3.5R and ϭ hIG 3.5RT. For this reason, it is also appropriate to compare our Redlich-Kwong free-jet results with the well established ideal gas ␥ ϭ 1.4 free-jet results. ␳ ϭ ␳ 1/2 The speed of sound c( ,T) (dP/d )s enters explicitly into the calculation as a stability constraint on the time step. In order to achieve the accuracy required, using any of our equa- ␳ tions of state, we numerically evaluated the derivative (dP/d )s at any ␳ and T by selecting a small ⌬␳ about the local ␳ and calculating the associated ⌬P for an isentropic process using both P(␳,T) and s(␳,T) equations of state.