|Home | About | Journals | Submit | Contact Us | Français|
Using liquid slugs as microreactors and microvessels enable precise control over the conditions of their contents on short-time scales for a wide variety of applications. Particularly for screening applications, there is a need for control of slug parameters such as size and composition. We describe a new microfluidic approach for creating slugs in air, each comprising a size and composition that can be selected individually for each slug. Two-component slugs are formed by first metering the desired volume of each reagent, merging the two volumes into an end-to-end slug, and propelling the slug to induce mixing. Volume control is achieved by a novel mechanism: two closed chambers on the chip are initially filled with air, and a valve in each is briefly opened to admit one of the reagents. The pressure of each reagent can be individually selected and determines the amount of air compression, and thus the amount of liquid that is admitted into each chamber. We describe the theory of operation, characterize the slug generation chip, and demonstrate the creation of slugs of different compositions. The use of microvalves in this approach enables robust operation with different liquids, and also enables one to work with extremely small samples, even down to a few slug volumes. The latter is important for applications involving precious reagents such as optimizing the reaction conditions for radiolabeling biological molecules as tracers for positron emission tomography.
The online version of this article (doi:10.1007/s10404-010-0617-0) contains supplementary material, which is available to authorized users.
Microfluidic technology has made it possible to perform biochemical assays and chemical reactions in extremely tiny volumes. A particularly successful approach has been to use microdroplets or liquid slugs as individual microreactors, within which the reaction conditions can be very precisely controlled on short-time scales (Song et al. 2006). Applications that exploit the unique properties of droplets and slugs have been reported extensively in the literature, including optimization of tracer synthesis conditions for positron emission tomography (Olma et al. 2008), protein crystallization (Chao et al. 2004; Lau et al. 2007; Zheng et al. 2003), single-cell isolation and analysis (Sakai et al. 2004; Sugiura et al. 2007), micro- and nano-particle synthesis (Hung et al. 2006; Li et al. 2008, 2009a; Nie et al. 2005; Nisisako et al. 2006; Shestopalov et al. 2004; Yen et al. 2005), biochemical assays (Srisa-Art et al. 2007), and small-scale chemical reactions (Frenz et al. 2008; Ota et al. 2009; Park et al. 2009). Droplets and slugs have also been useful for increasing the interfacial area between the droplet/slug phase and an intervening carrier phase for performing multi-phase processes such as gas–liquid reactions (Doku et al. 2005; Gunther et al. 2005; Wada et al. 2006), liquid–liquid extractions (Jensen et al. 2001; Kralj et al. 2005, 2007), and distillations (Hartman et al. 2009), among others.
Every application depends critically on techniques and technologies that accurately generate the desired droplets or slugs, and there has been a flurry of research activity in this area (Teh et al. 2008). Many efforts have focused on continuous flow technologies for the generation of monodisperse droplets of fixed size and composition (Hsiung et al. 2006; Kim et al. 2006), with multi-component droplets formed either by injecting laminar streams into a carrier flow (Tice et al. 2004) or by forming each component as separate droplets that are later fused by controlling the downstream flow (Hung et al. 2006; Liu et al. 2007; Utada et al. 2005). In both the cases, mixing occurs rapidly during transport. Control of droplet size and composition has been achieved by dynamically controlling flow rates (Collins and Lee 2007) or by the integration of active modulation by, e.g., electric fields (Ristenpart et al. 2009), temperature (Nguyen et al. 2007), mechanical chopping (Hsiung et al. 2006), and piezoactuation (Pan et al. 2007) into chips. Droplets and slugs are most frequently generated in an oil carrier phase in such continuous flow devices, but liquid-in-gas and gas-in-liquid slugs have also been demonstrated. Sophisticated screening assays have been performed by generating gradients of composition across many droplets (Laval et al. 2007; Zheng et al. 2003), or preparing sequences of precursor slugs in advance (Adamson et al. 2006; Hatakeyama et al. 2006; Li et al. 2006), but to the best of our knowledge, it remains a challenge to use these devices to efficiently (i.e., with low waste) generate droplets or slugs of independently selected composition when the total volume of reagents is small (e.g., tens to hundreds of slugs).
Another technology for creating droplets of controlled composition in air (or liquid) is digital microfluidics (Abdelgawad and Wheeler 2009; Chang and Yang 2007; Cho and Moon 2008), for example electrowetting-on-dielectric (EWOD) devices. Feedback control techniques have been demonstrated that can accurately dispense liquid droplets of precise volume, and these single-component droplets can then be merged and mixed using well-established on-chip operations (Cho et al. 2003; Gong and Kim 2008). Microfluidic devices containing microvalves provide an alternative approach for the generation of slugs with chips that are easier to fabricate and control. Like in EWOD devices, timing of droplet formation can be accurately controlled, improving device reliability and facilitating automation. Generation of single-component slugs in air has been demonstrated (Bontoux et al. 2006; Gu et al. 2004; Thorsen et al. 2002), as have manipulations of slugs including merging and mixing (Hansen et al. 2004; Lau et al. 2007; Sassa et al. 2008; Zeng et al. 2009). The use of mechanically acting microvalves to physically separate slugs from the incoming reagent streams reduces the dependence of operation of liquid properties such as surface tension, viscosity, electrowetting response, etc.
Building on the microvalve approaches, we have developed a novel device for the generation of two-component liquid slugs that permits selection of composition over a continuous range that uses dead-end filling to avoid reagent waste, and that can work with slug volumes down to the 10-nl volume range. For each slug, reagents are metered by partially filling two chambers, the volume of each liquid determined by controlling the compression of gas that is initially trapped in each chamber. The liquid volumes are merged to form an end-to-end slug of the desired size and composition that can then be mixed by transporting it. This slug generator increases the degree of control over slug formation and can be combined with other on-chip functions to perform a variety of biological, chemical, and physical applications. An advantage of using integrated microvalves is that the incoming reagents can very easily be synchronized and slugs can be correctly generated starting with the very first one. This enables the use of slugs to perform screening of reaction conditions, even when the total volumes are very small (e.g., down to a few slug volumes). This device is ideally suited for screening applications with rare or expensive samples, such as enabling the optimization of reaction conditions of 18F-labeled biomolecules as tracers for positron emission tomography (PET) to facilitate development and study of this rapidly emerging class of tracers (Liu et al. 2010).
The microfluidic slug generator (Fig. 1) was implemented as a two-layer polydimethylsiloxane (PDMS) chip according to common design and fabrication practices (Melin and Quake 2007). The upper layer of channels contains the fluidic channel network, consisting of reagent inlets, filling chambers, and slug output, while the lower layer contains microvalve control channels. Integrated “pinch valves” are created at each location where a control channel crosses directly below a fluid channel (Studer et al. 2004; Unger et al. 2000). The thin, elastic, PDMS membrane between the two channels can be actuated by applying pressure to the control channel, thus locally pinching off (closing) the fluid channel.
To fabricate the chip, photoresist-patterned silicon wafer molds for the fluid- and control-layers were first created by standard photolithographic techniques. Molds were pretreated by exposure to trimethylsilyl chloride (TMSCl) vapor for 10 min to facilitate mold release. Well-mixed PDMS prepolymer (GE RTV615, component ratio A/B = 5:1) was poured onto the fluid layer mold to give a 6-mm thick fluidic layer with approximately 60-μm channel depth. Another portion of PDMS prepolymer (GE RTV615, component ratio A/B = 20:1) was mixed and then spin-coated on the control layer mold at 1500 RPM for 60 s. The fluidic and control layers were cured at 80°C for 15 and 18 min, respectively, then aligned and then bonded by baking at 80°C for at least 6 h. Holes were then punched into the assembly to form ports connected to liquid inlets and outlets in the fluidic layer, and to valve control channels in the control layer. Strong adhesion of the chip to a clean glass microscope slide to seal the control channels was achieved by corona discharge treatment (Enercon Industries Corp., Dyne-A-Mite HP) (Eddings et al. 2008). The microfluidic device was baked at 80°C for 72 h to restore the intrinsic hydrophobicity of PDMS surfaces required for manipulation of aqueous samples.
Fluids were delivered to the chip through PTFE microbore tubing, connected via short stainless steel tubes inserted directly into the chip. Liquids A and B were manually primed up to the valves V4 and V6 on the chip (Fig. 2) using the vent channels. Microvalve control lines were filled with water. On-chip valves were actuated by pressurizing the corresponding control channel to 410 kPag via electronic solenoid valves (Series S070, SMC, Japan). The corresponding valve burst pressure was measured to be about 280 kPag, suggesting that fluid driving pressures up to about 210 kPag could be reliably used in the chip. All valves were automatically controlled through a data acquisition module (USB-4750, Advantech, USA) driven by a custom software program written in LabView (National Instruments, USA).
The sequence of valve states to operate the slug generator is shown in Fig. 2. The core of the chip consists of two adjacent chambers, A and B, of fixed size (about 16-nl each) that can be isolated from the surrounding channels by microvalves, providing a means for precise volumetric metering of reagents. In the starting step, Step 1, all valves are sealed. Step 2 is used to ensure a well-defined air pressure, Pinitial, initially inside each chamber. After the previous slug is ejected (as described below), the chambers contain air at pressure Pinitial = Pcarrier. Optionally, vent channels can be opened to release this pressure to atmosphere that establishes Pinitial = Patm. When used, venting was typically performed for tvent = 10 s to ensure sufficient time for pressures to equilibrate. Adjustment of Pinitial should be done immediately prior to Step 3 to prevent significant change in pressure due to gas permeation through the PDMS. In Step 3, two liquids, A and B, are filled into these chambers driven by pressures Pfill,A and Pfill,B. The volume loaded into each chamber depends on Pinitial, Pfill, and the filling time, tfill. Because the chambers are isolated from one another during the filling process, and the three parameters can be independently adjusted for the two chambers, the volume of each liquid can be individually controlled, providing a means to dynamically tune the size and composition of every slug generated. After isolation of the two liquid volumes (Step 4), the chamber contents are merged together in Step 5 to form a single end-to-end slug. The process of sealing off the two chambers after filling isolates the measured liquids from the inlet streams, therefore decoupling this slug formation (merging) process from the properties and driving pressures of the two incoming liquids. In Step 6, the slug is pushed out of the chambers to the chip outlet with carrier gas (nitrogen) at pressure Pcarrier. A slight narrowing of the downstream channel creates back-pressure that “cushions” the slug motion and prevents fragmentation of the slug that can occur when using relatively high Pcarrier. Once the slug is ejected from the filling chambers, preparation of the next slug can begin.
Mixing of slugs occurs rapidly by an internal recirculation pattern as it is transported along the downstream fluid path (Rhee and Burns 2008). In this chip, slugs are injected sequentially into tubing. In many applications it may be preferable to couple the slug generator directly to other on-chip components such as reaction channels, (de)multiplexers, storage chambers, or sensors.
The primary goal of this work is to develop an approach for reliable creation of two-component slugs with controlled composition. Since the composition is dictated by the amount of liquid filled into each chamber, we characterized the filling process as a function of adjustable parameters (Pinitial, Pfill, and tfill) to determine the best approach for controlling slug composition.
Filling occurs independently in the two chambers due to the presence of microvalve V5, and therefore experiments could be performed using only a single chamber. In each experiment, the fraction of the chamber filled with liquid (filling fraction, f) was estimated by the ratio of the “length” of the liquid-filled portion to the total length of the chamber at any given time. Length observations were made by recording video of chip operation using a CCD camera (BigC Dino-Lite). The slug generator could be paused at will, permitting sufficient time to capture clear images without sophisticated high-speed visualization. Unless otherwise indicated, each data point is an average of three measurements from different slugs. Error bars represent the standard deviation. The largest source of error in these measurements was uncertainty in the length, due to factors such as an asymmetric meniscus. Observed filling fraction under various conditions was compared with theoretical predictions.
The PDMS channels bulge when filling pressure is applied (Holden et al. 2003), so to make accurate estimates of volume filled, the chamber contents were sealed after filling for trelax = 5 s, during which time the filled portion was observed to “lengthen” as trapped air escaped through the PDMS and the channel relaxed to its original cross-section. This ensures that all length measurements are made with the same channel cross-section and thus can be directly compared even if filling pressures are different.
The microfluidic chip forms two-component slugs by merging the contents of two chambers containing reagents each with individually controlled volume. Partial dead-end filling of the chambers is studied in detail as a means to enable dynamic control of slug size and composition.
First, the temporal aspects of the filling process were studied. Using a single chamber, filling fraction of deionized water was measured as a function of filling time, tfill (Fig. 3). Filling was observed to proceed in two distinct phases with very different time scales. Initially, within tens of milliseconds, the chamber fills to a certain fraction (shown later to depend on filling pressure, Pfill). The second phase is much slower (tens of seconds) and appears approximately linear. The combination of very high and very low slopes makes it difficult to use tfill as a means to control the fraction filled over a wide range. However, the insensitivity to time (e.g., 4.4% per second for the first 1 s in Fig. 3) does have an advantage: if tfill is chosen within the second phase, filling fraction will not be sensitive to small inaccuracies in filling time (e.g., tens to hundreds of milliseconds) that may arise due to factors such as electronics delays and finite valve response time.
We chose to operate our device with a fixed filling time, selected just past the transition between the two phases, i.e., tfill = 100 ms. Keeping the filling time short permits higher slug generation rates, and also ensures the filling fraction is close to its value at the end of the fast phase, which can be predicted with good accuracy (see below). The 100-ms filling time was chosen based on data collected with deionized water. For other liquids, it may be necessary to choose a slightly longer filling time to ensure the slow filling phase is reached. An example of filling with glycerol is described in the Supplementary Information.
Filling appears to occur by rapid compression of the trapped gas initially in the chamber (fast phase), followed by permeation of this pressurized gas out through the PDMS matrix (slow phase). In the fast compression phase, the final steady-state “plateau” that is reached can be predicted from the ideal gas law, independent of any fluid properties or device geometry. We ignore the dynamics of how quickly the steady state is achieved, which would involve factors that affect fluid movement, such as viscosity, geometry of channel and interconnect, filling pressure, and total liquid volume. (This issue is addressed in the Supplementary Information where it is shown that by choosing a slightly longer tfill these dynamics can be ignored.)
The fraction of the chamber filled at the end of the fast compression phase can be modeled as a function of Pfill and Pinitial. Assume that a chamber of volume V0 is initially filled with air at pressure Pinitial. When filling begins, this gas rapidly compresses to a steady-state pressure Pfill. Assuming there is negligible loss of gas or temperature change, the ideal gas law should hold, and the final volume of compressed air is:
The remaining volume of the chamber is now filled with liquid from the inlet channel. Thus for the given pressures, the fraction filled with liquid at the end of the compression phase is expected to be:
This simple model has no fitted parameters.
Equation 2 predicts that with a fixed filling time selected just after the end of the compression phase, the filling fraction depends on both the filling pressure Pfill and the initial pressure, Pinitial. Experimentally, the expected dependence on both the pressures is seen (see Fig. 4a, b), and comparison with the ideal gas model (solid line) after conversion of gauge to absolute pressures of Eq. 2 shows good agreement at low Pfill. However, there is increasing discrepancy between experimental data and the model at higher filling pressures, presumably due to the expansion of the PDMS channel under pressure. Modeling of this effect is discussed in the next section.
Both curves are monotonic and have moderate slope; thus, either of these two pressures, Pfill or Pinitial, can be manipulated to control the filling fraction of liquids, and thus the composition of slugs. Linear dependence on Pinitial makes Pinitial preferable in principle, but because it is difficult to accurately maintain pressure inside a PDMS chip due to the high permeability of the PDMS material, we chose to use Pfill as a control variable instead. Pinitial was fixed at 0 kPag by venting the chambers prior to filling to create the next slug. Negative pressure was considered and would provide a greater range of filling fractions, but it cannot be as accurately maintained as atmospheric pressure (0 kPag).
Two different filling pressures, Pfill,A and Pfill,B, for the two incoming liquids can be used to tune dynamically the degree of filling of each, thereby determining the composition and size of the liquid slugs generated. Using a fast electronic pressure regulator to change driving pressures, it is feasible to select the size and the composition of every slug individually. Over the filling pressure ranges used, we achieved fractions filled covering the range from about 0.1 to 0.8 as shown in Fig. 4, enabling slugs composed of A/B = 1:8 to 8:1 to be created.
The model of Eq. 2 assumes that the initially trapped air is compressed to a pressure equal to the filling pressure, Pfill. However, if the liquid–air interface is not flat during filling, an additional capillary pressure is generated (Geerken et al. 2007; Li et al. 2009b). The result of this correction would be a lateral shift in plots such as Fig. 4a. We estimate the magnitude, based on the properties of water and PDMS as well as the device geometry, to be on the order of 1 kPa or less. The capillary pressure would therefore have only very small effect on filling fraction (for the range of filling pressures used in this study), and therefore is ignored.
The low elastic modulus of PDMS results in substantial deformation of channel dimensions at high internal pressure (Gervais et al. 2006; Holden et al. 2003). The channel cross-section expands during filling due to the application of pressure Pfill. The volume fraction filled still follows the ideal gas model, but because the cross-section is larger, the chamber actually contains more liquid than if the channel did not expand. The magnitude of the discrepancy depends on the change in cross-sectional size which increases with on filling pressure. This is evident in the data in Fig. 4a and b that exceed the simple ideal gas model that neglects the expansion.
To ensure that all data points were comparable, so that length of the filled portion reflects the volume fraction of the original-sized chamber, we measured filled lengths after permitting chambers to relax to original dimensions. During this delay, trapped air permeates out of the chamber through the PDMS, gradually reducing the pressure and thus the deformation (see Fig. 4c). As the channel dimensions decrease, the liquid is redistributed to fill a longer length of the channel.
To model the discrepancy between actually amount filled and the prediction of Eq. 2, we estimate the equilibrium deformation of the chamber under the filling pressure, Pfill, and make a correction to the expected length fraction that accounts for the extra volume filled due to this deformation. The geometry of this problem can be roughly approximated as a thick-walled elastic cylinder that is internally pressurized, where the chamber/channel radius represents the inner radius of the cylinder, the bulk PDMS of the chip represents the wall of the cylinder, and the outside of the chip represents the outer radius of the cylinder. Though chips in reality are not cylinders centered around microchannels we in fact find that the functional form of this model has good agreement with the experimental data.
The model predicts (Wedgwood 1928) that the channel experiences a radial strain that is linearly related to the internal pressure P:
where k is an expression that depends on the elastic modulus and Poisson’s ratio of the PDMS as well as geometry of the “cylinder.”
The cross-sectional area of the channel therefore enlarges to , where A0 is the original cross-section. The ideal gas model predicts that a fraction f of this larger chamber will be filled. We can express the total volume filled as . After relaxation to the original channel size A0, the measured length fraction can therefore be predicted as:
where f is the prediction of the ideal gas model from Eq. 2.
This model, plotted in Figs. 4 and and55 (dotted lines), has one fitted parameter, k, and shows good agreement with the data when k = 0.00065 kPag−1. The dependence on, and sensitivity to, the material and geometrical properties should be investigated in more detail to ensure operational parameters remain constant from chip to chip. As a preliminary test, we performed experiments with several different chips of varying ages and observed only about 1% discrepancy in chamber filling fractions (data not shown).
We have shown above how to control the filling fraction of a single chamber by varying Pfill. Under the conditions used in this article, the fraction filled depends only on the ideal gas law, and an adjustment for the expansion of PDMS channels under pressure that depends on PDMS properties and geometry. There should be no dependence on fluid properties such as viscosity, contact angles, total volume, etc., other than brief transient effects during the initial part of the filling process.
For a desired slug size and ratio of components, it is possible to determine (from data in Fig. 4 or Eq. 4) what filling pressures should be used. We applied our microfluidic chip to the creation of slugs with arbitrarily selected compositions in the range 8:1–1:8. Figure 5 illustrates an example where several different slug compositions were generated by maintaining a constant filling pressure of liquid A (water), Pfill,A = 140 kPag, while varying the filling pressure of liquid B (water), Pfill,B from 20 to 170 kPag. Compositions access in this experiment ranged from about A/B = 4.5–0.9.
Using food dye in water, end-to-end slugs were qualitatively observed to be well mixed within a few slug lengths of movement through the channel downstream of the slug generator. We did not quantitatively analyze the mixing because this principle has been studied by numerous researchers (Rhee and Burns 2008).
An important application of the ability to controllably generate tiny mixtures in this fashion is in optimizing reaction conditions with precious reagents. As an example, ongoing research efforts in our laboratory are investigating the use of the slug generator for radiolabeling of biological molecules for use as molecular imaging tracers in positron emission tomography (PET). In these reactions, the labeling yield is very sensitive to parameters such as concentrations, pH, and type/sequence of biomolecule (Liu et al. 2010). To maximize the yield, both to conserve biomolecule and to achieve superior quality images, one generally must perform trial-and-error optimization. At the conventional scale (100 s of μl) this would consume enormous amounts of biomolecule and so optimization is rarely done, resulting often in poor yield. However, in this microfluidic device reaction mixtures could be prepared on demand in truly tiny volumes (10 s of nl), each of which can be reacted and analyzed by existing off-chip analytical tools to guide the optimization process. The space of reaction parameters could then be explored quite extensively with only minimal reagent consumption.
Unlike continuous flow devices, where the process of droplet and slug formation is affected by properties and flow rates of the incoming streams as well as the contents of the downstream channel (thus depending on the history of operation), formation of slugs in our device is insulated from these effects. Isolating the chambers from the liquid inlets after filling decouples slug formation (merging) from the properties and driving pressures of the two incoming liquids. In addition, isolation of the chambers from the downstream channel decouples the filling and slug formation process from the history of operation. These properties are what enable instant selection of new slug composition at any time without a transition period that wastes reagents. They also permit slug generation to start and stop on demand, making it easier to synchronize slug formation and integrate with other subsystems in an automated fashion.
Despite these advantages, the approach presented here may not be suitable for certain applications, due to limitations in operating speed and fluid compatibility. Operating speed is limited by the microvalve response time and is significantly slower than the high slug generation rates that can be achieved in some continuous approaches; however, it is still quite acceptable for a wide range of screening applications, especially those involving relatively small numbers of slugs or where the slug analysis time is long. Fluid compatibility is limited by the use of PDMS which swells in many organic solvents (Lee et al. 2003). It may be possible to expand the range of applications by fabricating similar devices with inert materials such as perfluoropolyether (Rolland et al. 2004) or hybrids of glass and Teflon (Grover et al. 2008).
Improvements in accuracy of volume ratios might be achieved using non-permeable, non-elastic materials for the filling chamber. This would improve the ability to accurately maintain Pinitial, and would also eliminate the channel expansion that currently introduces some dependence of slug generation on the mechanical properties of the chip material.
Our lab is currently investigating a different approach for improving ratio accuracy that retains the advantages of fabricating chips from PDMS. This approach will be based on the concept of quantized filling fractions, achieved by the use of valves to subdivide each chamber into several subchambers that are each either completely filled or empty. We are also exploring the extension of these concepts to accurate generation of slugs composed of three or more reagents.
We have demonstrated and characterized a microfluidic device for the robust generation of multi-component liquid-in-air slugs for applications such as screening of reaction conditions involving precious reagents. The device is a proof of concept demonstrating that microfluidic platforms with integrated microvalves can produce slug flows with the advantage of programmable size and composition of slugs. Selected volumes of two reagents are loaded into separate chambers and then merged together into a single end-to-end slug that can be efficiently mixed by transport along the downstream microchannel. The slug is separated from the input reagent streams by mechanically actuating valves, decoupling the slug formation process from the influence of fluid parameters. Slugs are also isolated from the downstream channel during formation. Two-component slug generation using our approach should therefore be more predictable and robust than other methods of slug formation over a range of reagents. A simple model of the volume-control process was reported and found to be in excellent agreement with experimental data. A key advantage of the chip reported here is that it can operate reliably even with very small total volumes of the two reagents (<1 μl) and with minimal waste of reagents. Accurate slug formation begins with the very first slug and can be continued until the very last one. Size and composition can be programmed individually for each slug, with no transition period during which slugs are malformed, such as the stabilization time needed in continuous flow approaches. Straightforward extensions to our design could enable the formation of programmable slugs comprising three or more reagents.
This study was supported in part by the National Cancer Institute (Nanosystems Biology Cancer Center, Grant number U54CA119347-01), the Industry-University Cooperative Research Program (UC Discovery Grant number bio07-10665) and a project development grant from the UCLA Center for In Vivo Imaging in Cancer Biology (National Institutes of Health Grant number 5P50CA086306). The authors thank Dirk Williams and Darin Williams for assistance in designing and fabricating custom chip connectors.
Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
Kan Liu and Yi-Chun Chen contributed equally to this work.
Clifton Kwang-Fu Shen, Phone: 310-983-3058, Fax: 310-206-8975, Email: ude.alcu.tendem@nehsk.
R. Michael van Dam, Phone: 310-206-6507, Fax: 310-206-8975, Email: ude.alcu.tendem@madnavm.