Two-dimensional insertable separation tool (TWIST) for flow confinement in spatial separations

Spatial comprehensive two-dimensional liquid chromatography (xLC×xLC) may be an efficient approach to achieve high peak capacities in relatively short analysis times, thanks to parallel second-dimension separations [1,2]. A key issue to reach the potential of xLC×xLC is to achieve adequate flow control and confinement of the analytes to the desired regions, i.e. confinement in the first-dimension direction and subsequently homogeneous flow in the second dimension. To achieve these goals we propose the TWIST concept (TWo-dimensional Insertable Separation Tool), a modular device that includes an internal first-dimension (1D) part that is cylindrical and rotatable. This internal part features a series of throughholes, each of which is perpendicular to the direction of the 1D flow. The internal part is inserted in the cylindrical casing of the external part. The internal diameter of the casing is marginally larger than the external diameter of the internal part. The external part also comprises a flow distributor and seconddimension (2D) channels. During the 1D injection and development, the channel is placed in a position D-printing omputational fluid dynamics where the through-holes are facing the wall of the external part, such that the liquid remains confined within the 1D channel. Thereafter, to realize the transfer to the second dimension (2D injection), the 1D channel is rotated, so that the holes of the internal part are aligned with the holes on the external part, allowing a transversal flow of the 2D mobile phase from the distributor through the 1D channel and eventually into the 2D area. © 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license


a b s t r a c t
Spatial comprehensive two-dimensional liquid chromatography ( x LC× x LC) may be an efficient approach to achieve high peak capacities in relatively short analysis times, thanks to parallel second-dimension separations [1,2]. A key issue to reach the potential of x LC× x LC is to achieve adequate flow control and confinement of the analytes to the desired regions, i.e. confinement in the first-dimension direction and subsequently homogeneous flow in the second dimension. To achieve these goals we propose the TWIST concept (TWo-dimensional Insertable Separation Tool), a modular device that includes an internal first-dimension ( 1 D) part that is cylindrical and rotatable. This internal part features a series of throughholes, each of which is perpendicular to the direction of the 1 D flow. The internal part is inserted in the cylindrical casing of the external part. The internal diameter of the casing is marginally larger than the external diameter of the internal part. The external part also comprises a flow distributor and seconddimension ( 2 D) channels. During the 1 D injection and development, the channel is placed in a position where the through-holes are facing the wall of the external part, such that the liquid remains confined within the 1 D channel. Thereafter, to realize the transfer to the second dimension ( 2 D injection), the 1 D channel is rotated, so that the holes of the internal part are aligned with the holes on the external part, allowing a transversal flow of the 2 D mobile phase from the distributor through the 1 D channel and eventually into the 2

Introduction
Comprehensive two-dimensional liquid chromatography (LC × LC) is indispensable for the characterization of very complex samples [3]. Greatly enhanced peak capacities relative to conventional one-dimensional (1D) LC may be obtained by LC × LC, which can be effectively realized if the two separation dimensions are sufficiently different (i.e. highly orthogonal).
One premise for successful comprehensive operation is that the entire first-dimension effluent must be transferred and subjected to the second-dimension separation. This requires the chromatographer to establish a compromise between the first-and second-dimension column dimensions and flow rates and the modulation time, which implies a sacrifice in performance. The need to compromise can be avoided with a perfectly operated spatial x LC× x LC system. A suitable format for spatial separations may be realized through microfluidic devices [2,4]. However, perfect oper- * Corresponding author. ation requires rigorous confinement of the flow of mobile phase and analytes in the desired ( 1 D or 2 D) direction. Incomplete confinement will greatly affect the separation efficiency.
The development of microfluidic devices is typically a stepwise process of design and prototyping. Actual prototyping can be a time-consuming and cumbersome task [5]. By using computational fluid dynamics (CFD), designs can be theoretically established and tested. Satisfactory designs are then prototyped and the resulting experimental performance can be used to enhance the design further. Previous CFD studies have been performed on flow distribution, 1 D injection volumes and channel discretization in the second dimension [1,6,7]. To facilitate rapid and easy prototyping, 3D-printing methods have been adopted. Stereo-lithography provides a high degree of accuracy and consistency with the original design. In the present study a novel flow-confinement concept (two-dimensional insertable separation tool or TWIST) for spatial comprehensive two-dimensional liquid chromatography ( x LC× x LC) devices is presented.

Computational fluid dynamics (CFD) studies
For CFD simulations, ANSYS Workbench Fluids and Structures Academic Package (ANSYS, Pennsylvania, PA, USA; version 17.1) was used. The cases were solved as 3D domains and for the entire geometry to be simulated.
The examined devices consisted of three main parts, viz. the flow distributor, the 1 D channel and the 2 D domain. To investigate (lack of) flow confinement, a solution of dye in water corresponding to three 1 D channel volumes was introduced to the 1 D channel at a flow rate of 5 mL min −1 . During the 1 D injection the 2 D inlet and outlet were kept closed.
All dimensions were chosen in accordance with our 3D printing capabilities (see Section 2.2 below). Simulations were conducted with both empty 1 D channels and channels filled with a porous structure. The computations involving empty 1 D channels were in compliance with previously fabricated and studied devices, in which iso-electric focusing was used as a separation method in the first dimension [4,8]. The cases simulated with a porous 1 D channel represented the presence of a monolith as stationary phase. The permeability of 1 D and 2 D porous structures was 1.7·10 −13 m 2 [9].
All cases were meshed in a similar manner, with size inflation at the distributor, body sizing at the 1 D channel and edge sizing with divisions at the 2 D area. In the 2 D domain bias was imposed in order to have more resolution close to the wall and to the 1 D to 2 D transition area. These modifications were made to enhance the accuracy of the CFD results.

Designing and 3D-printing
The design process was facilitated by the commercial package Autodesk Inventor (Autodesk, San Rafael, CA, USA). The proposed design is depicted in Fig. 1 as an assembly. It consists of two parts: an internal (grey) and an external (blue) one. The internal part, also shown for clarity in the insert, is the channel in which the 1 D separation takes place. Two diametric series of through-holes are created to allow a perpendicular flow to pass through the 1 D channel during transfer to the second dimension. The external part comprises a flow distributor (top), the 1 D channel casing in which the internal part is inserted, and a series of parallel 2 D channels (bottom).
The examined device was fabricated through 3D-printing using a Digital Light Processing (DLP) Asiga Pico 2 HD (385 nm). Printing orientation and settings were optimized for high resolution. After 3D-printing, post processing of the parts was necessary. This included sonication and flushing the channels with 2-propanol and nitrogen to remove any uncured resin. When all the undesirable material was removed, the parts were inserted in a Pico Flash UV chamber (type DR-301C, 36 W, 365 nm, 3DXS Germany) and cured between 30 and 90 min (depending on the part). To make the devices connectable, straight threads (#10-32 UNC, major diameter 4.83 mm, thread pitch 0.794 mm) were created using a hand tap, whereas the conical part at the end of the connection area was already incorporated into the print design. In order to connect the tubing to the flow distributor inlet, the tubing was inserted 2 mm in the inlet channel and glued with optical glue. The connection for the 2 D inlet was not included in the design to stay within the surface area of the printer's build platform and the chosen printing orientation (the part was printed horizontally). The final connections were watertight at the pressures needed for testing.

Chemicals and materials
Asiga PlasClear V2 resin was purchased from 3DXS (Erfurt, Germany). 2-Propanol (Biosolve BV, Valkenswaard, The Netherlands) was used during the post-processing of the printed parts, as well as during the flow tests. The PME Natural Food Color -Red (product nr: PFC1022, www.deleukstetaartenshop.nl) and PME Natural Food Color -Blue (supplied by local source) were used during testing. Nitrogen used during the post-processing of the printed parts was supplied by Praxair to a laboratory gas-supply network. Optical glue was supplied by a local source.

Flow testing
Two set of experiments had to be performed; a flowconfinement investigation during the 1 D injection and transfer from the 1 D to the 2 D. Flow-confinement tests were performed on the fabricated device. The device was empty and a mixture of dye dissolved in water was injected in the 1 D channel for flow visualization. Different flow rates in the range of 0.5 to 5 mL/min were studied. For the second investigation, the 1 D channel was rotated and water was injected at 1.5 ml/min from the inlet of the flow distributor. The flow profiles were recorded with a Canon EOS 1300D camera. The experiments were performed without a holder or additional sealing.

Results and discussion
All designs presented in this study consisted of three main parts, viz. a flow distributor, a 1 D channel and a 2 D domain. In these type of devices either a monolith is assumed to be present in both the 1 D and 2 D domains, or only in the 2 D domain, leaving the 1 D channel empty.

Flow confinement
Flow confinement is necessary during the 1 D step. Leakage during this step (from the first-dimension channel to the flow distributor or to the second-dimension channels) can undermine both the first-and the second-dimension separations. In Fig. 2 exemplary results are depicted of 1 D injections into devices without any confinement measures and assuming a monolithic packing in the 1 D channel (left) or an empty 1 D channel (right). The desired outcome is to have the dye present only in the 1 D channel (in high concentrations, i.e. red in the figure). Both cases in Fig.2 are seen to result in excessive amounts of dye in other compartments of the device.
In case of an empty 1 D channel (right) leakage is observed mainly towards the flow distributor area. In the case with a porous 1 D channel (left), dye penetrates both to the distributor and the 2 D area. The dramatic dye distribution in the latter case can be understood by realizing that the flow in the 1 D direction creates a pressure gradient from left to right in the figure. This gradient makes the liquid follow the path of the least flow resistance on its way to the outlet, making a detour through the 2 D flow distributor before exiting along the exit of the 1 D channel. One way to enhance flow confinement would be to keep the 1 D channel empty (right panel) and to create constrictions, such as monolithic frits, at the outlets of the distributor, minimizing the leakage of dye to the 2 D flow distributor (result not shown). However, an empty 1 D channel leaves few separation options other than IEF. A different design is needed that allows a stationary phase to be present in the 1 D channel, while achieving flow confinement and effective flow control.
As a solution to the above problem, we propose a concept wherein the 1 D separation takes place in a channel with a cylindrical external geometry [patent pending, nr. EP18184801.1]. This channel can be inserted in a cylindrical housing in the 2 D device. Both the 1 D channel and the housing contain through-holes. During the operation of the device for the 1 D separation the through-holes of the chamber are not aligned with the through-holes of the 1 D channel. During the subsequent second-dimension separation, the through-holes of the chamber are aligned with the through-holes of the insertable channel, allowing a perpendicular flow through the 1 D channel. A great additional advantage is that the insertable 1 D channel can easily be replaced, allowing different stationary phases to be used for the 1 D separation. The 1 D separation may also be performed in a different housing (off-line), for example one that allows higher pressures for the flow to pass through the full length of the channel.

Flow testing
Some leakage (through the first hole in the internal cylinder) was observed at very low flow rates (0.5 mL/min) at the inlet side of the channel probably related to the residence time of the dye solution. Also, some leakage (between the internal and the external parts) was observed at high flow rates (5 mL/min) or upon prolonged flushing. No leakage was observed during standard operation at 1.5 or 2 mL/min. These results were obtained without any sealing in place. Leakages can be reduced by incorporating sealing in the device e.g. by incorporating frits or membranes in the holes and adding O-rings or sleaves. In Fig. 3 the three main steps of operation are presented. Initially the 1 D injection takes place, while the through-holes on the 1 D channel are not aligned with the distributor and the 2 D channels (a). Afterwards the 1 D channel is rotated to achieve the desired alignment (b). Here, the liquid present in the dead zone inside the holes can be observed. Dead zones can be minimized, for example by placing a frit or a membrane. Finally, frame (c) depicts the successful emptying of the 1 D channel to the 2 D, proving the principle and correct operation of the device.

Conclusions
Rigorous flow confinement in the 1 D channel is required for optimal operation of spatial comprehensive two-dimensional liquid chromatography ( x LC× x LC) devices. To achieve this a twodimensional insertable separation tool (TWIST) is proposed. This concept allows confinement of the flow and independent separations to be performed in the different dimensions. A prototype was made using 3D-printing technology. Further research is required for device and material optimization, incorporation of stationary phases and for performing actual separations. Furthermore, use of an external holder could offer stability during the rotation of the 1 D channel and accuracy for the required alignment. Finally, with some modifications, the TWIST concept may provide an attractive option to realize flow confinement in spatial three-dimension liquid chromatography.