NOTE

A Simple Confined Impingement Jets Mixer for FlashNanoprecipitation

JING HAN,1 ZHENGXI ZHU,1 HAITAO QIAN,2 ADAM R. WOHL,2 CHARLES J. BEAMAN,2 THOMAS R. HOYE,2

CHRISTOPHER W. MACOSKO1

1Department of Chemical Engineering and Materials Science, College of Science and Engineering, University of Minnesota,Minneapolis, Minnesota 55455

2Department of Chemistry, College of Science and Engineering, University of Minnesota, Minneapolis, Minnesota 55455

Received 14 March 2012; revised 26 May 2012; accepted 15 June 2012

Published online in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.23259

ABSTRACT: Johnson and Prud′homme (2003. AICHE J 49:2264–2282) introduced the con-fined impingement jets (CIJ) mixer to prepare nanoparticles loaded with hydrophobic com-pounds (e.g., drugs, inks, fragrances, or pheromones) via flash nanoprecipitation (FNP). Wehave modified the original CIJ design to allow hand operation, eliminating the need for a sy-ringe pump, and we added a second antisolvent dilution stage. Impingement mixing requiresequal flow momentum from two opposing jets, one containing the drug in organic solvent andthe other containing an antisolvent, typically water. The subsequent dilution step in the newdesign allows rapid quenching with high antisolvent concentration that enhances nanoparticlestability. This new CIJ with dilution (CIJ-D) mixer is a simple, cheap, and efficient device toproduce nanoparticles. We have made 55 nm diameter $-carotene nanoparticles using the CIJ-Dmixer. They are stable and reproducible in terms of particle size and distribution. We have alsocompared the performance of our CIJ-D mixer with the vortex mixer, which can operate at un-equal flow rates (Liu et al., 2008. Chem Eng Sci 63:2829–2842), to make $-carotene-containingparticles over a series of turbulent conditions. On the basis of dynamic light scattering mea-surements, the new CIJ-D mixer produces stable particles of a size similar to the vortex mixer.Our CIJ-D design requires less volume and provides an easily operated and inexpensive toolto produce nanoparticles via FNP and to evaluate new nanoparticle formulation. © 2012 WileyPeriodicals, Inc. and the American Pharmacists Association J Pharm SciKeywords: CIJ mixer; nanoparticles; flash nanoprecipitation; drug delivery; $-carotene; mix-ing; nanotechnology; particle size

INTRODUCTION

Nanoparticles have recently received enormous at-tention as a drug delivery tool.1–5 Flash nanoprecip-itation (FNP) is a simple technique that is used toprepare polymeric nanoparticles with a high load-ing of hydrophobic compounds, including drugs.6–8

As shown in Figure 1,9 a hydrophobic drug andan amphiphilic block copolymer [e.g., polyethyleneglycol-b-polylactic acid (PEG-b-PLA)] are codissolvedin a water-miscible organic solvent [e.g., tetrahydro-furan (THF)], which is then impinged at high velocityagainst an antisolvent (water) to create turbulentmixing and high supersaturation. The supersatu-

Correspondence to: Christopher W. Macosko (Telephone: +612-625-0092; Fax: +612-626-1686; E-mail: macosko@umn.edu)Journal of Pharmaceutical Sciences© 2012 Wiley Periodicals, Inc. and the American Pharmacists Association

ration promotes coprecipitation of the hydrophobicdrug and the hydrophobic block of the copolymer toform nanoparticles.10–13 Mixing and precipitation oc-cur within milliseconds inside the small internal mix-ing chamber.

Johnson and Prud′homme6 first described FNP us-ing a confined impingement jets (CIJ) mixer. In thisdesign, a syringe pump was used to drive two op-posing liquid streams (a and b in Fig. 1) at highvelocity into the mixing chamber. Prud′homme andothers6,7,14–17,33 have used this device to successfullymake a variety of nanoparticles. The CIJ design wasinspired by the simple T mixer that is commonly usedto mix liquids or act as a chemical reactor.18,19 Tmixers have also been used to mix monomers andoligomers for reaction injection molding.20–23

To avoid reducing the mixing efficiency (via onestream backing up the other), the two streams in

JOURNAL OF PHARMACEUTICAL SCIENCES 1

2 HAN ET AL.

Figure 1. Schematic of flash nanoprecipitation process.9

the CIJ mixer must operate at near equal momen-tum. In practice, this requires approximately a 1:1volume flow rate, limiting the highest achievable su-persaturation level. To operate at unequal flow ratios,Prud′homme and coworkers13,24 developed a multi-inlet vortex mixer. In this design, the mixing cham-ber is connected to four inlets and the liquid streamsmeet at an angle as opposed to the head-on impinge-ment that is characteristic of the CIJ mixer. The vor-tex mixer can be applied to a wide range of solventratios and materials. However, the device is time con-suming to clean, and digitally programmed syringepumps are usually required to control the inlet flowrates.25 In terms of operation, cleaning, and cost, theCIJ mixer is preferred. Nevertheless, the vortex mixeris able to achieve higher levels of supersaturation.

Here, we report a modified CIJ design, the CIJ withdilution (CIJ-D) mixer, that is simpler than the orig-inal Johnson and Prud′homme design, yet overcomesthe limitation of 1:1 solvent ratio. To emphasize itssimpler design and easier handling with equivalentfunction, we have compared the average size of $-carotene nanoparticles made using our CIJ-D mixerversus the vortex mixer.

DESIGN

Figure 2 shows the CIJ-D design. Two features dis-tinguish it from the original CIJ mixer: hand opera-tion and an antisolvent dilution stage. By using rela-tively small, low-friction syringes, turbulent flow canbe achieved with simple, rapid hand motion, eliminat-ing the need for syringe pumps. The small sample sizeand easy operation made the CIJ-D ideal for screen-ing candidate formulations. A metal plate connectsthe two syringes to ensure simultaneous actuation.To increase the supersaturation, the outlet streamfrom the CIJ-D chamber immediately flows directlyinto a large volume of water.

The design dimensions for the CIJ mixing cham-ber was recommended by Johnson and Prud′homme.6

Figure 2. The CIJ-D mixer, hand operated with subse-quent dilution.

Two pathways lead to a small chamber with a con-fined volume, where the organic solvent and water im-pinge to create turbulence. A ratio of entrance channellength to diameter (L/d) = 6.1 was used to insure sta-ble jets. The ratio of chamber height above the inletnozzle to diameter is H = 0.8D, and the length to di-ameter ratio H + Z = 2.0D was held constant to main-tain geometric similarity upon scale-up.6 The cham-ber volume is 25:L. The outlet tube runner lengthshould be at least 10 times the outlet diameter K/δ >

10, to create a pressure drop and to ensure that thechamber is filled with liquid during impingement.

Figure 3 shows the dimensions of our CIJ-D cham-ber. The main body of the CIJ mixer was made of high-density polyethylene, with two inlets and adapters(IDEX Health & Science, P604, Middleboro, Mas-sachusetts) fitted with threaded syringes, and oneoutlet adapter (IDEX Health & Science, P205-X). Twoadditional side openings resulted when horizontal jetpathways were drilled during manufacturing. Theseports are sealed with threaded plugs (IDEX Health &Science, P203-X) during FNP experiments, but can beopened for thorough cleaning.

A typical procedure for making nanoparticles isas follows. $-Carotene was used as the generichydrophobic molecule (log P = 15.5, ACD modelfrom www.emolecules.com). PEG-b-PLA (molecu-lar weight: 5000-b-10,000 Da26) was used as the

JOURNAL OF PHARMACEUTICAL SCIENCES DOI 10.1002/jps

A SIMPLE CONFINED IMPINGEMENT JETS MIXER FOR FLASH NANOPRECIPITATION 3

Figure 3. Dimensions of the CIJ-D mixer made of high-density polyethylene.

amphiphilic block copolymer. Twenty-five milligrams$-carotene and 25 mg PEG-b-PLA were dissolved in2.5 mL of THF and transferred to a 3 mL plasticsyringe (Kendall, Tyco Healthcare, Mansfield, Mas-sachusetts). Deionized water (2.5 mL) was loaded ina second 3 mL plastic syringe. A metal plate wasplaced on the top of the syringes to ensure simul-taneous actuation, which occurred over a period ofapproximately 5 s. The comixed stream was immedi-ately diluted into a bottom reservoir preloaded with45 mL of deionized water. The final composition ofthe 50 mL dispersion was THF/H2O = 5:95, contain-ing 0.1 wt % of nanoparticles. Mass-average particlesize and size distribution was determined by dynamiclight scattering (DLS) (ZetaPALS, Brookhaven In-struments, Holtsville, New York; diode laser BI-DPSSwavelength of 659 nm, round cuvette) using regular-ized positive exponential sum (REPES)0 method,27,28

immediately after nanoparticles were prepared fol-lowing the procedure of Zhu.9 The light intensity cor-relation function was collected at 25◦C and a scatter-ing angle of 90◦.

PERFORMANCE

To test the CIJ-D mixer, nanoparticles were pre-pared following the typical procedure given above.DLS gave mass-average diameter of 38 nm with poly-dispersity of 1.6 and standard deviation of 6 nm forthree measurements on three separately mixed sam-ples. The particles were stable for several weeks. Inother block copolymer and model drug studies, wehave also used PEG-b-PLA and PEG-b-poly(lactic-co-glycolic acid) with a series of molecular weight(5000-b-10,000 Da, 5000-b-15000 Da, etc.) to makenanoparticles loaded with $-carotene. Zhu9 used hy-drocortisone, paclitaxel, betulin, and their derivativesto test the capability of the CIJ-D mixer for prepar-

ing polymeric nanoparticles. Chow et al.29 assessedthe CIJ-D mixer by making curcumin nanoparticles.Prud′homme and coworkers30 have used the CIJ-Dmixer making polystyrene nanoparticles. They weresuccessfully produced, reproducible in terms of size.

As mentioned above, the dilution stage allows highsupersaturation while maintaining the 1:1 flow ratioof the impinging streams. In order to demonstrate theimportance of the dilution stage, four groups of par-ticles were made following the procedure describedin the section Design but with differing dilution. Forthis study, we used pure $-carotene (no block copoly-mer) to make particles. These particles show goodshort-term (∼4 h) stability because of the slightly neg-ative surface charge31 as judged by zeta potential (ζ)measurements.321

In group 1, nanoparticles were made followingthe recommended procedure, immediate dilution into45 mL water, THF/H2O = 0.05. The resulting mass-average size was 55 nm. In group 2, nanoparticleswere made without dilution. They were unstable, andaggregated to micron size in seconds. In group 3, dilu-tion was delayed from ∼10 ms (residence time in themixer and outlet tube) to 5 s. This also resulted in un-stable microparticles (>1:m). In group 4, nanoparti-cles were made with immediate (10 ms) water dilutionbut less than 45 mL (e.g., 5, 10, and 25 mL or THF/H2O = 0.33, 0.20, and 0.091). The particle sizes fromgroup 4 are shown in Figure 4. None of them produced55 nm nanoparticles, in contrast to the usual dilu-tion, THF/H2O = 0.05. Instead, all were much bigger,approximately 1400, 200, and 167 nm, respectivelyand eventually unstable. These experiments showthat immediate dilution with a significant amountof water is indispensable to produce small and stablenanoparticles.

We also compared the sizes of these pure $-carotenenanoparticles made by the CIJ-D mixer to those madeby a vortex mixer. Particle size was varied by varying

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES

4 HAN ET AL.

Figure 4. Mass-average diameter of $-carotene particles versus THF/water ratio in group 4(Re ≈ 1750).

the flow rate, Q. This was accomplished simply bychanging hand velocity during mixing. Mixer effi-ciency is typically correlated with Reynolds number(Re), the ratio of inertial force to viscous force. The Rewas calculated by:

Re = Inertial force(= ρV2)Viscous force(= μV/d)

= ρVdμ

= ρQcμA

(1)

In our case, Re was calculated by accumulatingmultiple streams when using the CIJ-D or vortexmixer.

Re =n∑

i=1

Rei = dA

n∑

i=1

ρiQi

μi= 4

Bd

n∑

i=1

ρiQi

μi(2)

where ρ is the density of the fluid (kg/m3); V is themean fluid velocity (s); d is the stream inlet diameter;μ is the viscosity of the fluid (kg/m s); A is the pipecross-sectional area (m2), which in our case was thesame for all inlets of a given mixer; and Qi is thevolumetric flow rate (m3/s) of the ith inlet stream.

For example, in CIJ mixing, n = 2 for two streams,ρ is 1.0 × 103 kg m−3 for H20 or 8.89 × 102 kg m−3

for THF at room temperature, μ is 1.0 × 10−3 Pa s atroom temperature for H2O or 4.8 × 10−4 Pa s for THF,and d is 5 × 10−4 m for the CIJ-D mixer.

The vortex mixer has four inlets, each with di-ameter d = 1.45 × 10−3 m. For all four streams, weassumed μ = 1.0 × 10−3 Pa s and ρ = 1.0 × 10−3

kg m−3 at room temperature because 90%H2O/

Figure 5. Mass-average size of $-carotene particles (dm) versus Reynolds number (Re). Par-ticles were made using both the vortex mixer and CIJ-D mixer. The Re was calculated usingEq. 2.

JOURNAL OF PHARMACEUTICAL SCIENCES DOI 10.1002/jps

A SIMPLE CONFINED IMPINGEMENT JETS MIXER FOR FLASH NANOPRECIPITATION 5

10%THF has a similar kinematic viscosity to that ofwater.

The mass-average diameters shown in Figure 5 areaverages of three measurements on three separatelymixed samples prepared under the same conditions.The vortex mixing data were taken from Zhu.9 Theerror in the calculated Re comes from variation ofinjection time. In vortex mixing, errors are minor be-cause the injection time is mechanically controlled bysyringe pumps, but it is relatively large in CIJ-D mix-ing due to the uncertainty in timing of the hand mo-tion. At the highest Re, the injection time of the CIJ-Dmixer was 4.2 ± 0.3 s. 2.5 mL of solvents in two 3 mLsyringes were used; thus, the flow rate was 1.2 × 10−6

m3/s. Considering this flow rate and the tubing di-mensions (1 mm inside diameter and 15 mm long), theresidence time in the chamber and outlet tube beforedilution was approximately 10 ms. For Re < 1000, asyringe pump was used with the CIJ-D mixer becausehand operation cannot be controlled well enough tocreate steady mixing over minutes.

The results in Figure 5 show good agreement be-tween the new CIJ-D design and the vortex mixer.Because of the rapid injection, the CIJ-D can reachRe > 4000, twice the limit for the vortex mixer, andparticle size of 30 nm, half the smallest size from thevortex mixer. At the same Re, particle size for theCIJ-D is slightly smaller.

A potential problem with hand operation of theCIJ-D mixer is the start-up of flow. The Re at the be-ginning of impingement will be lower because of thetime required to accelerate the syringes. The sameproblem occurs with syringe pumps, but because theimpingement time is longer, it is possible to discardthe first part of the product. This start-up transientcould lead to broader particle size distribution; how-ever, we observed broad size distribution in both theCIJ-D mixer and the vortex mixer. For example, atRe ≈ 1750 (shown in Fig. 5), DLS gave averaged sizepolydispersity indices9 of 0.5−0.8 for both samples.

CONCLUSION

A simple modification of Johnson and Prud′homme’sCIJ mixer has been used to make stable and repro-ducible nanoparticles. The addition of a dilution stageafter mixing results in higher levels of supersatura-tion, overcoming the limitation of equal volume ra-tios required in the original CIJ design. For our $-carotene particles, we found that the dilution neededto be rapid, <5 s after mixing, and extensive, greaterthan fivefold. Hand-operated impingement with smallsyringes creates sufficient turbulent mixing, ideal forFNP of hydrophobic compounds with diblock copoly-mers to form kinetically trapped, sterically stabi-lized nanoparticles. With these modifications, the newmixer, called CIJ-D to emphasize the addition of a di-

lution stage, uses very small volumes and is easy tooperate and inexpensive, making it more effective forrapid screening of small quantities of new materi-als, via FNP, compared with the alternative mixersand other methods. It is especially attractive for eval-uation of new drug formulations for their ability toproduce nanoparticles.

ACKNOWLEDGMENTS

We appreciate design help from Carl Johnson of theUniversity of Minnesota Physics and Astronomy Ma-chine Shop. This work was supported by the Univer-sity of Minnesota Futures Grant Program and theNational Institutes of Health (EB011671).

REFERENCES

1. Feng SS, Chien S. 2003. Chemotherapeutic engineering: Appli-cation and further development of chemical engineering prin-ciples for chemotherapy of cancer and other diseases. ChemEng Sci 58:4087–4114.

2. Ferrari M. 2005. Cancer nanotechnology: Opportunities andchallenges. Nat Rev Cancer 5:161–171.

3. Faraji AH, Wipf P. 2009. Nanoparticles in cellular drug deliv-ery. Bioorg Med Chem 17:2950–2962.

4. Horn D, Rieger J. 2001. Organic nanoparticles in the aque-ous phase-theory, experiment and use. Angew Chem Int Ed40:4330–4361.

5. Wang X, Yang L, Chen Z, Shin DM. 2008. Application of nan-otechnology in cancer therapy and imaging. CA Cancer J Clin58:97–110.

6. Johnson BK, Prud′homme RK. 2003. Chemical process-ing and micromixing in confined impinging jets. AICHE J49:2264–2282.

7. Johnson BK, Prud′homme RK. 2003. Flash NanoPrecipitationof organic actives and block copolymers using a confined im-pinging jets mixer. Aust J Chem 56:1021–1024.

8. Johnson BK, Prud′homme RK. 2004. Process and apparatusesfor preparing nanoparticle compositions with amphiphiliccopolymers and their use. Patent US0091546 A1.

9. Zhu ZX. 2010. Polymer stabilized nanosuspension via flashnanoprecipitation: Nanoparticle formation, formulation andstability. Ph.D. Thesis.Minnesota: University of Minnesota.

10. Cheng CY, Vigil RD, Fox RO. 2010. A competitive aggrega-tion model for flash nanoprecipitation. J Colloid Interface Sci351:330–342.

11. Gavi E, Marchisio DL, Barresi AA. 2007. CFD modeling andscale-up of confined impinging jet reactors. Chem Eng Sci62:2228–2241.

12. Liu Y, Fox RO. 2006. CFD predictions for chemical processingin a confined impinging-jets reactor. AICHE J 52:731–744.

13. Liu Y, Cheng CY, Liu Y, Prud′homme RK, Fox RO. 2008.Mixing in a multi-inlet vortex mixer (MIVM) for flash nano-precipitation. Chem Eng Sci 63:2829–2842.

14. Akbulut M, Ginart P, Gindy ME, Theriault C, Chin KH,Soboyejo W, Prud′homme RK. 2009. Generic method of prepar-ing multifunctional fluorescent nanoparticles using flashnanoprecipitation. Adv Funct Mater 19:718–725.

15. Johnson BK, Prud′homme RK. 2003. Mechanism for rapidself-assembly of block copolymer nanoparticles. Phys Rev Lett91:1183021–1183024.

16. Lince F, Marchisio DL, Barresi AA. 2009. Smart mixers andreactors for the production of pharmaceutical nanoparticles:Proof of concept. Chem Eng Res Des 87:543–549.

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES

6 HAN ET AL.

17. Marchisio DL, Rivautella L, Barresi AA. 2006. Design andscale-up of chemical reactors for nanoparticles precipitation.AICHE J 52:1877–1887.

18. Hunter TG, Nash AW. 1935. Liquid–liquid extraction systems.Ind Eng Chem 27:836–845.

19. Treybal RE. 1963. Liquid extraction. 2nd ed. New York:McGraw-Hill.

20. Lee LJ, Ottino JM, Ranz WE, Macosko CW. 1980. Impinge-ment mixing in reaction injection molding. Polym Eng Sci20:868–874.

21. Macosko, CW. 1989. RIM, fundamentals of reaction injectionmolding. New York: Hanser.

22. Nguyen LT, Suh NP. 1986. Processing of polyurethane/polyester interpenetrating polymer networks by reaction in-jection molding: Part 2. Mixing at high Reynolds num-bers and impingement pressures. Polym Eng Sci 26:799–842.

23. Tucker CL, Suh NP. 1980. Mixing for reaction injectionmolding. I. Impingement mixing of liquids. Polym Eng Sci20:875–886.

24. Gindy ME, Panagiotopoulos AZ, Prud′homme RK. 2008. Com-posite block copolymer stabilized nanoparticle: Simultaneousencapsulation of organic actives and inorganic nanostructure.Langmuir 24:83–90.

25. D′addio SM. 2011. Flash nanoprecipitation video. Accessed, at:http://www.princeton.edu/research/news/princetoninvention/video/index.xml. (Last accessed: 12/13/11)

26. Qian HT, Wohl AR, Crow JT, Macosko CW, Hoye TR. 2011.A strategy for control of “random” copolymerization of lac-tide and glycolide: Application to synthesis of PEG-b-PLGA

block polymers having narrow dispersity. Macromolecules44(18):7132–7140.

27. Jakes J. 1988. Testing of the constrained regularizationmethod of inverting laplace transform on simulated very widequasielastic light-scattering auto-correlation functions. CzechJ Phys 38(12):1305–1316.

28. Jakes J. 1995. Regularized positive exponential sum (REPES)program—A way of inverting laplace transform data obtainedby dynamic light scattering. Collect Czech Chem Commun60(11):1781–1797.

29. Chow S, Sun C, Macosko CW, Chow A. 2011. Comparativeassessment of the performance between confined impinge-ment jet and multi-inlet vortex mixers for controlled produc-tion of curcumin nanoparticles. In American Association ofPharmaceutical Scientists 2011 Meeting. October 23–27, 2011;McCormick Place, Chicago, Illinois.

30. Zhang C, Pansare VJ, Prud′homme RK, Priestley RD. 2012.Flash nanoprecipitation of polystyrene nanoparticles. SoftMatter 8:86–93.

31. Mordi RC, Walton JC, Burton GW, Hughes L, Ingold KU,Lindsay DA, Moffatt DJ. 1993. Oxidative-degradation of $-Carotene and $-8′-carotenal. Tetrahedron 49:911–928.

32. Zhu ZX, Margulis-Goshen K, Magdassi S, Talmon Y,Macosko CW. 2010. Polyelectrolyte stabilized drug nanopar-ticles via flash nanoprecipitation: A model study with $-carotene. J Pharm Sci 99:4295–4306.

33. Wohl AR, Kalscheuer S, Lee HS, Han J, McCormick A,Macosko CW, Panyam J, Hoye TR. A silicate ester prodrugstrategy for improving the therapeutic index of paclitaxel. JAm Chem Soc (Inpreparation).

JOURNAL OF PHARMACEUTICAL SCIENCES DOI 10.1002/jps