On the moral responsibility of military robots

This article discusses mechanisms and principles for assignment of moral responsibility to intelligent robots, with special focus on military robots. We introduce the concept autonomous power as a new concept, and use it to identify the type of robots that call for moral considerations. It is furthermore argued that autonomous power, and in particular the ability to learn, is decisive for assignment of moral responsibility to robots. As technological development will lead to robots with increasing autonomous power, we should be prepared for a future when people blame robots for their actions. It is important to, already today, investigate the mechanisms that control human behavior in this respect. The results may be used when designing future military robots, to control unwanted tendencies to assign responsibility to the robots. Independent of the responsibility issue, the moral quality of robots’ behavior should be seen as one of many performance measures by which we evaluate robots. How to design ethics based control systems should be carefully investigated already now. From a consequentialist view, it would indeed be highly immoral to develop robots capable of performing acts involving life and death, without including some kind of moral framework.     

A mechanism for sharing best practices between university technology transfer offices

Research has shown that university technology transfer offices (TTOs) learn through experimentation and failure, and by sharing these experiences with others. There are many barriers to successfully sharing the best practice between TTOs. The Maturity Model (MM) created by Secundo et al. (Meas Bus Excell, 20:42–54, 2016) provides a means by which the performance of a TTO can be better understood to allow for effective sharing of best practices. The aim of this study is to improve and validate the MM to formalize a mechanism through which best practices can be identified and shared between TTOs. This was accomplished by testing the MM in 54 TTOs across Europe and the United Kingdom. Findings regard several improvements of the intangible indicators and the maturity levels of the MM. This research improves the rigor of the MM and formalizes its application as a mechanism for sharing best practices through the Improved MM.     

A new solution to the gamer’s dilemma

Luck (2009) argues that gamers face a dilemma when it comes to performing certain virtual acts. Most gamers regularly commit acts of virtual murder, and take these acts to be morally permissible. They are permissible because unlike real murder, no one is harmed in performing them; their only victims are computer-controlled characters, and such characters are not moral patients. What Luck points out is that this justification equally applies to virtual pedophelia, but gamers intuitively think that such acts are not morally permissible. The result is a dilemma: either gamers must reject the intuition that virtual pedophelic acts are impermissible and so accept partaking in such acts, or they must reject the intuition that virtual murder acts are permissible, and so abstain from many (if not most) extant games. While the prevailing solution to this dilemma has been to try and find a morally relevant feature to distinguish the two cases, I argue that a different route should be pursued. It is neither the case that all acts of virtual murder are morally permissible, nor are all acts of virtual pedophelia impermissible. Our intuitions falter and produce this dilemma because they are not sensitive to the different contexts in which games present virtual acts.     

Social robots,fiction, and sentimentality

I examine the nature of human-robot pet relations that appear to involve genuine affective responses on behalf of humans towards entities, such as robot pets, that, on the face of it, do not seem to be deserving of these responses. Such relations have often been thought to involve a certain degree of sentimentality, the morality of which has in turn been the object of critical attention (Sparrow in Ethics Inf Technol 78:346–359, 2002; Blackford in Ethics Inf Technol 14:41–51, 2012). In this paper, I dispel the claim that sentimentality is involved in this type of relations. My challenge draws on literature in the philosophy of art and in cognitive science that attempts to solve the so called paradox of fictional emotions, i.e., the seemingly paradoxical way in which we respond emotionally to fictional or imaginary characters and events. If sentimentality were not at issue, neither would its immorality. For the sake of argument, however, I assume in the remaining part of the paper that sentimentality is indeed at play and bring to the fore aspects of its badness or viciousness that have not yet been discussed in connection with robot pets. I conclude that not even these aspects of sentimentality are at issue here. Yet, I argue that there are other reasons to be worried about the wide-spread use of ersatz companionship technology that have to do with the potential loss of valuable, self-defining forms of life.     

Community of practices,knowledge transfer,and ERP project (ERPP)

In this research, we are interested in ERP systems which are common information repositories that are aimed at matching the knowledge, practices, and skills that drive the organization in the best possible way. Can the cognitive and hierarchical models coexist within the same project? What is the impact of ERP on the interconnection between communities? To answer these questions, we rely in particular on the work of Levina and Vaast (MIS Quarterly 29(2):335–363, 2005), which underlines that the modes of interaction between CPs must be mediated by the activation of boundary objects and/or the mobilization of boundary spanners. Finally, this leads us to discriminate between two types of ERPPs (hierarchical/cognitive) and to underline the role of the switch in the ERPP success.     

Society-in-the-loop: programming the algorithmic social contract

Recent rapid advances in Artificial Intelligence (AI) and Machine Learning have raised many questions about the regulatory and governance mechanisms for autonomous machines. Many commentators, scholars, and policy-makers now call for ensuring that algorithms governing our lives are transparent, fair, and accountable. Here, I propose a conceptual framework for the regulation of AI and algorithmic systems. I argue that we need tools to program, debug and maintain an algorithmic social contract, a pact between various human stakeholders, mediated by machines. To achieve this, we can adapt the concept of human-in-the-loop (HITL) from the fields of modeling and simulation, and interactive machine learning. In particular, I propose an agenda I call society-in-the-loop (SITL), which combines the HITL control paradigm with mechanisms for negotiating the values of various stakeholders affected by AI systems, and monitoring compliance with the agreement. In short, ‘SITL = HITL + Social Contract.’     

Exploring Social Theory as a Framework for Social and Cultural Measurements of the Information Society

Using Layder's domain theory (1997) Layder, D. 1997. Modern social theory:Key debates and new directions, London: UCL Press.  [Google Scholar] as an analytical framework, this article shows how the information society can be measured through various levels of society. Layder's notions of psychobiography, situated activity, social setting, and contextual resources help identify cultural and social indicators for understanding changes in the information society. With the help of empirical indicators for each domain, this article uses the case of Estonia to show that there is often more to the information society than what is captured by traditional measures. This article calls for a context-sensitive approach, which takes into consideration social and cultural indicators. Measurements from all four domains are necessary for understanding the complexity of information-society-related issues.     

Polyphosphonium‐based bipolar membranes for rectification of ionic currents

Bipolar membranes (BMs) have interesting applications within the field of bioelectronics, as they may be used to create non-linear ionic components (e.g., ion diodes and transistors), thereby extending the functionality of, otherwise linear, electrophoretic drug delivery devices. However, BM based diodes suffer from a number of limitations, such as narrow voltage operation range and/or high hysteresis. In this work, we circumvent these problems by using a novel polyphosphonium-based BM, which is shown to exhibit improved diode characteristics. We believe that this new type of BM diode will be useful for creating complex addressable ionic circuits for delivery of charged biomolecules.Combined electronic and ionic conduction makes organic electronic materials well suited for bioelectronics applications as a technological mean of translating electronic addressing signals into delivery of chemicals and ions.¹ For complex regulation of functions in cells and tissues, a chemical circuit technology is necessary in order to generate complex and dynamic signal gradients with high spatiotemporal resolution. One approach to achieve a chemical circuit technology is to use bipolar membranes (BMs), which can be used to create the ionic equivalents of diodes^{2, 3, 4, 5} and transistors.^{6, 7, 8} A BM consists of a stack of a cation- and an anion-selective membrane, and functions similar to the semiconductor PN-junction, i.e., it offers ionic current rectification^{9, 10} (Figure ​(Figure1a).1a). The high fixed charge concentration in a BM configuration make them more suited in bioelectronic applications as compared to other non-linear ionic devices, such as diodes constructed from surface charged nanopores¹¹ or nanochannels,¹² as the latter devices typically suffers from reduced performance at elevated electrolyte concentration (i.e., at physiological conditions) due to reduced Debye screening length.¹³ However, a unique property of most BMs, as compared to the electronic PN-junction and other ionic diodes, is the electric field enhanced (EFE) water dissociation effect.^{10, 14} This occurs above a threshold reverse bias voltage, typically around −1 V, as the high electric field across the ion-depleted BM interface accelerates the forward reaction rate of the dissociation of water into H⁺ and OH⁻ ions. As these ions migrate out from the BM, there will be an increase in the reverse bias current. The EFE water dissociation is a very interesting effect and is commonly used in industrial electrodialysis applications,¹⁵ where highly efficient water dissociating BMs are being researched.¹⁶ Also, BMs have also been utilized to generate H⁺ and OH⁻ ions in lab-on-a-chip applications.^{2, 17} However, the EFE water dissociation effect diminishes the diode property of BMs when operated outside the ±1 V window, which is unwanted in, for instance, chemical circuits and addressing matrices for delivery of complex gradients of chemical species. The effect can be suppressed by incorporating a neutral electrolyte inside the BM,^{10, 18} for instance, poly(ethylene glycol) (PEG).^{2, 6, 7} However, as previously reported,² the PEG volume will introduce hysteresis when switching from forward to reverse bias, due to its ability to store large amounts of charges. This was circumvented by ensuring that only H⁺ and OH⁻ are present in the diode, which recombines into water within the PEG volume. Such diodes are well suited as integrated components in chemical circuits for pH-regulation, due to the in situ formed H⁺ and OH⁻, but are less attractive if, for instance, other ions, e.g., biomolecules, are to be processed or delivered in and from the circuit. Furthermore, a PEG electrolyte introduces additional patterning layers, making device downscaling more challenging. This is undesired when designing complex, miniaturized, and large-scale ionic circuits. Thus, there is an interest in BM diodes that intrinsically do not exhibit any EFE water dissociation, are easy to miniaturize, and that turn off at relatively high speeds. It has been suggested that tertiary amines in the BM interface are important for efficient EFE water dissociation,^{19, 20, 21} as they function as a weak base and can therefore catalyze dissociation of water by accepting a proton. For example, anion-selective membranes that have undergone complete methylation, converting all tertiary amines to quaternary amines, shows no EFE water dissociation,¹⁹ although this effect was not permanent, as the quaternization was reversed upon application of a current. Similar results were found for anion-selective membranes containing alkali-metal complexing crown ethers as fixed charges.²¹ Also, EFE water dissociation was not observed or reduced in BMs with poor ion selectivity, for example, in BMs with low fixed-charge concentration⁵ or with predominantly secondary and tertiary amines in the anion-selective membrane,²² as the increased co-ion transport reduces the electric field at the BM interface. However, due to decreased ion selectivity, these membranes show reduced rectification. In this work, we present a non-amine based BM diode that avoids EFE water dissociation, enables easy miniaturization, and provides fast turn-off speeds and high rectification.Open in a separate windowFigure 1(a) Ionic current rectification in a BM. In forward bias, mobile ions migrate towards the interface of the BM. The changing ion selectivity causes ion accumulation, resulting in high ion concentration and high conductivity. At high ion concentration, the selectivity of the membranes fails (Donnan exclusion failure), and ions start to pass the BM. In reverse bias, the mobile ions migrate away from the BM, eventually giving a zone with low ion concentration and low conductivity. Reverse bias can cause EFE water dissociation, producing H⁺ and OH^- ions. (b) Chemical structures of PSS, qPVBC, and PVBPPh₃. (c) The device used to characterize the BMs and the BM1A, BM2A, and BM1P designs. The BM interfaces are 50 × 50 μm.An anion-selective phosphonium-based polycation (poly(vinylbenzyl chloride) (PVBC) quaternized by triphenylphospine, PVBPPh₃) was synthesized and compared to the ammonium-based polycation (PVBC quaternized by dimethylbenzylamine, qPVBC) previously used in BM diodes² and transistors,^{7, 8} when included in BM diode structures together with polystyrenesulfonate (PSS) as the cation-selective material (Figure ​(Figure1b).1b). Three types of BM diodes were fabricated using standard photolithography patterning (Figure ​(Figure1c),1c), either with qPVBC (BM1A and BM2A) or PVBPPh₃ (BM1P) as polycation and either with (BM2A) or without PEG (BM1A and BM1P). Poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) electrodes covered with aqueous electrolytes were used to convert electronic input signals into ionic currents through the BMs, according to the redox reaction PEDOT⁺:PSS⁻ + M⁺ + e⁻ ↔ PEDOT⁰ + M⁺:PSS⁻.The rectifying behavior of the diodes was evaluated using linear sweep voltammetry (Figure ​(Figure2).2). The BM1A diode exhibited an increase in the reverse bias current for voltages lower than −1 V, a typical signature of EFE water dissociation,^{10, 14} which decreased the current rectification at ±4 V to 6.14. No such deviation in the reverse bias current was observed for BM2A and BM1P, which showed rectification ratios of 751 and 196, respectively. In fact, for BM1P, no evident EFE water dissociation was observed even at −40 V (see inset of Figure ​Figure2).2). Thus, the PVBPPh₃ polycation allows BM diodes to operate at voltages beyond the ±1 V window with maintained high ion current rectification.Open in a separate windowFigure 2Linear sweep voltammetry from −4 to +4 V (25 mV/s) for the BM diodes. The inset shows BM1P scanning from −40 V to +4 V (250 mV/s).The dynamic performance of the diodes was tested by applying a square wave pulse from reverse bias to a forward bias voltage of 4 V with 5–90 s pulse duration (Figure ​(Figure3).3). To access the amount of charge injected and extracted during the forward bias and subsequent turn off, the current through the device was integrated. For BM2A, we observed that the fall time increased with the duration of the forward bias pulse. This hysteresis is due to the efficient storage of ions in the large PEG volume, with no ions crossing the BM due to Donnan exclusion failure.² As a result, during the initial period of the return to reverse bias, a large amount of charge needs to be extracted in order to deplete the BM. After a 90 s pulse, 90.6% of the injected charge during the forward bias was extracted before turn-off. This may be contrasted with BM1P, where the fall time is hardly affected by the pulse duration, and the extracted/injected ratio is only 15.4% for a 90 s pulse. For this type of BM, the interface quickly becomes saturated with ions during forward bias, leading to Donnan exclusion failure and transport of ions across the BM.⁴ Thus, less charge needs to be extracted to deplete the BM, allowing for faster fall times and significantly reduced hysteresis.Open in a separate windowFigure 3Switching characteristics (5, 10, 20, 30, 60, or 90 s pulse) and ion accumulation (at 90 s pulse) of the BM2A and BM1P diodes. BM1A showed similar characteristics as BM1P when switched at ±1V (see supplementary material).²⁴Since the neutral electrolyte is no longer required to obtain high ion current rectification over a wide potential range, the size of the BM can be miniaturized. This translates into higher component density when integrating the BM diode into ionic/chemical circuits. A miniaturized BM1P diode was constructed, where the interface of the BM was shrunk from 50 μm to 10 μm. The 10 μm device showed similar IV and switching characteristics as before (Figure ​(Figure4),4), but with higher ion current rectification ratio (over 800) and decreased rise/fall times (corresponding to 90%/–10% of forward bias steady state) from 10 s/12.5 s to 4 s/4 s. Since the overlap area is smaller, a probable reason for the faster switching times is the reduced amount of ions needed to saturate and deplete the BM interface. In comparison to our previous reported low hysteresis BM diode,² this miniaturized polyphosphonium-based devices shows the same rise and fall times but increased rectification ratio.Open in a separate windowFigure 4(a) Linear sweep voltammetry and (b) switching performance of a BM1P diode with narrow junction.In summary, by using polyphosphonium instead of polyammonium as the polycation in BM ion diodes the EFE water dissociation can be entirely suppressed over a large operational voltage window, supporting the theory that a weak base, such as a tertiary amine, is needed for efficient EFE water dissociation.^{17, 18} As no additional neutral layer at the BM interface is needed, ion diodes that operate outside the usual EFE water dissociation window of ±1 V can be constructed using less active layers, fewer processing steps and with relaxed alignment requirement as compared to polyammonium-based devices. This enables the fabrication of ion rectification devices with an active interface as low as 10 μm. Furthermore, the exclusion of a neutral layer improves the overall dynamic performance of the BM ion diode significantly, as there is less hysteresis due to ion accumulation. Previously, the hysteresis of BM ion diodes has been mitigated by designing the diode so that only H⁺ and OH⁻ enters the BM, which then recombines into water.² Such diodes also show high ion current rectification ratio and switching speed but are more complex to manufacture, and their application in organic bioelectronic systems is limited due to the H⁺/OH⁻ production. By instead using the polyphosphonium-based BM diode, reported here, we foresee ionic, complex, and miniaturized circuits that can include charged biomolecules as the signal carrier to regulate functions and the physiology in cell systems, such as in biomolecule and drug delivery applications, and also in lab-on-a-chip applications.     

Flow manipulation and cell immobilization for biochemical applications using thermally responsive fluids

A flow redirection and single cell immobilization method in a microfluidic chip is presented. Microheaters generated localized heating and induced poly(N-isopropylacrylamide) phase transition, creating a hydrogel that blocked a channel or immobilized a single cell. The heaters were activated in sets to redirect flow and exchange the fluid in which an immobilized cell was immersed. A yeast cell was immobilized in hydrogel and a 4′,6-diamidino-2-phenylindole (DAPI) fluorescent stain was introduced using flow redirection. DAPI diffused through the hydrogel and fluorescently labelled the yeast DNA, demonstrating in situ single cell biochemistry by means of immobilization and fluid exchange.The ability to control microfluidic flow is central to nearly all lab-on-a-chip processes. Recent developments in microfluidics either include microchannel based flow control in which microvalves are used to control the passage of fluid,¹ or are based on discrete droplet translocation in which electric fields or thermal gradients are used to determine the droplet path.^{2, 3} Reconfigurable microfluidic systems have certain advantages, including the ability to adapt downstream fluid processes such as sorting to upstream conditions and events. This is especially relevant for work with individual biomolecules and high throughput cell sorting.⁴ Additionally, reconfigurable microfluidic systems allow for rerouting flows around defective areas for high device yield or lifetime and for increasing the device versatility as a single chip design can have a variety of applications.Microvalves often form the basis of flow control systems and use magnetic, electric, piezoelectric, and pneumatic actuation methods.⁵ Many of these designs require complicated fabrication steps and can have large complex structures that limit the scalability or feasability of complex microfluidic systems. Recent work has shown how phase transition of stimuli-responsive hydrogels can be used to actuate a simple valve design.⁶ Beebe et al. demonstrated pH actuated hydrogel valves.⁷ Phase transition of thermosensitive poly(N-isopropylacrylamide) (PNIPAAm) using a heater element was demonstrated by Richter et al.⁸ Phase transition was also achieved by using light actuation by Chen et al.⁹ Electric heating has shown a microflow response time of less than 33 ms.¹¹ Previous work¹⁰ showed the use of microheaters to induce a significant shift in the viscosity of thermosensitive hydrogel to block microchannel flow and deflect a membrane, stopping flow in another microchannel. Additionally, Yu et al.¹² demonstrated thermally actuated valves based on porous polymer monoliths with PNIPAAm. Krishnan and Erickson¹³ showed how reconfigurable optically actuated hydrogel formation can be used to dynamically create highly viscous areas and thus redirect flow with a response time of  ～ 2?s. This process can be used to embed individual biomolecules in hydrogel and suppress diffusion as also demonstrated by others.^{15, 16} Fiddes et al.¹⁴ demonstrated the use of hydrogels to transport immobilized biomolecules in a digital microfluidic system. While the design of Krishnan and Erickson is highly flexible, it requires the use of an optical system and absorption layer to generate a geometric pattern to redirect flow.This paper describes the use of an array of gold microheaters positioned in a single layer polydimethylsiloxane (PDMS) microfluidic network to dynamically control microchannel flow of PNIPAAm solution. Heat generation and thus PNIPAAm phase transition were localized as the microheaters were actuated using pulse width modulation (PWM) of an applied electric potential. Additionally, hydrogel was used to embed and immobilise individual cells, exchange the fluid parts of the microfluidic system in order to expose the cells to particular reagents to carry out an in situ biochemical process. The PDMS microchannel network and the microheater array are shown in Figure ​Figure11.Open in a separate windowFigure 1A sketch of the electrical circuit and a microscope image of the gold microheaters and the PDMS microchannels. The power to the heaters was modulated with a PWM input through a H-bridge. For clarity, the electrical circuit for only the two heaters with gelled PNIPAAm is shown (H1 and V2). There are four heaters (V1-V4) in the “vertical channels” and three heaters (H1-H3) in the “horizontal” channel.The microchannels were fabricated using a patterned mould on a silicon wafer to define PDMS microchannels, as described by DeBusschere et al.¹⁷ and based on previous work.¹⁰ A 25 × 75 mm glass microscope slide served as the remaining wall of the microchannel system as well as the substrate for the microheater array. The gold layer had a thickness of 200 nm and was deposited and patterned using E-beam evaporation and photoresist lift-off.²¹ The gold was patterned to function as connecting electrical conductors as well as the microheaters.It was crucial that the microheater array was aligned with an accuracy of  ～ 20μm with the PDMS microchannel network for good heat localization. The PDMS and glass lid were treated with plasma to activate the surface and alignment was carried out by mounting the microscope slide onto the condenser lens of an inverted microscope (TE-2000 Nikon Instruments). While imaging with a 4× objective, the x, y motorized stage aligned the microchannels to the heaters and the condenser lens was lowered for the glass substrate to contact the PDMS and seal the microchannels.Local phase transition of 10% w/w PNIPAAm solution in the microchannels was achieved by applying a 7 V potential through a H-bridge that received a PWM input at 500 Hz which was modulated using a USB controller (Arduino Mega 2650) and a matlab (Mathworks) GUI. The duty cycle of the PWM input was calibrated for each microheater to account for differences in heater resistances (25?Ω to 52?Ω) due to varying lengths of on-chip connections and slight fabrication inconsistencies, as well as for different flow conditions during device operation. Additionally, thermal cross-talk between heaters required decreasing the PWM input significantly when multiple heaters were activated simultaneously. This allowed confining the areas of cross-linked PNIPAAm to the microheaters, allowing the fluid in other areas to flow freely.By activating the heaters in sets, it was possible to redirect the flow and exchange the fluid in the central area. Figure ​Figure22 demonstrates how the flow direction in the central microchannel area was changed from a stable horizontal flow to a stable vertical flow with a 3 s response time, using only PNIPAAm phase transition. Constant pressures were applied to the inlets to the horizontal channel and to the vertical channels. Activating heaters V1-4 (Figure ​(Figure2,2, left) resulted in flow in the horizontal channel only. Likewise, activating heaters H1 and H2 allowed for flow in the vertical channel only. In this sequence, the fluid in the central microchannel area from one inlet was exchanged with fluid from the other inlet. Additionally, by activating heater H3, a particle could be immobilised during the exchange of fluid as shown in Figure ​Figure33 (top).Open in a separate windowFigure 2Switching between fluid from the horizontal and the vertical channel using hydrogel activation and flow redirection with a response time of 3 s. A pressure of 25 mbar was applied to the inlet of the horizontal channel and a pressure of 20 mbar to the vertical channel. The flow field was determined using particle image velocimetry, in which the displacement of fluorescent seed particles was determined from image pairs generated by laser pulse exposure. Processing was carried out with davis software (LaVision).Open in a separate windowFigure 3A series of microscope images near heater H3 showing: (1a)-(1c) A single yeast cell captured by local PNIPAAm phase transition and immobilized for 5 min before being released. (2a) A single yeast cell was identified for capture by embedding in hydrogel. (2b) The cell as well as the hydrogel displayed fluorescence while embedded due to the introduction of DAPI in the surrounding region. (2c) The diffusion of DAPI towards the cell as the heating power of H3 is reduced after 15 min, showing a DAPI stained yeast cell immobilized.Particle immobilisation in hydrogel and fluid exchange in the central area of the microfluidic network were used to carry out an in situ biochemical process in which a yeast cell injected through one inlet was stained in situ with a 4′,6-diamidino-2-phenylindole (DAPI) solution (Invitrogen), which attached to the DNA of the yeast cell.¹⁸ A solution of yeast cells with a concentration of 5 × 10⁷cells/ml suspended in a 10% w/w PNIPAAm solution was injected through the horizontal channel. A solution of 2μg/l DAPI in a 10% w/w PNIPAAm solution was injected through the vertical channel. A single yeast cell was identified and captured near the central heater, and by deactivating the heaters in the vertical channel, DAPI solution was introduced in the microchannels around the hydrogel. After immobilising the cell for 15 min, the heater was deactivated, releasing the cell in the DAPI solution. This process is shown in Figure ​Figure33 (bottom). The sequence of the heater activation and deactivation in order to immobilize the cell and exchange the fluid is outlined in the supplementary material.²¹Eriksen et al.¹⁵ demonstrated the diffusion of protease K in the porous hydrogel matrix,¹⁹ and it was therefore expected that DAPI fluorescent stain (molecular weight of 350 kDa, Ref. ²⁰) would also diffuse. DAPI diffusion is shown in Figure 3(2b) in which the yeast cell shows fluorescence while embedded in the hydrogel. The yeast cell was released by deactivating the central heater and activating all the others to suppress unwanted flow in the microchannel. As a result, the single cell was fully immersed in the DAPI solution. Immobilization of a single cell allows for selection of a cell that exhibits a certain trait and introduction of a new fluid while maintaining the cell position in the field of view of the microscope such that a biochemical response can be imaged continuously.In summary, a microfluidic chip capable of local heating was used to induce phase transition of PNIPAAm to hydrogel, blocking microchannel flow, and thereby allowing for reconfigurable flow. Additionally, the hydrogel was used to embed and immobilise a single yeast cell. DAPI fluorescent stain was introduced using flow redirection, and it stained the immobilized cell, showing diffusion into the hydrogel. The versatile design of this microfluidic chip permits flow redirection, and is suitable to carry out in situ biochemical reactions on individual cells, demonstrating the potential of this technology for forming large-scale reconfigurable microfluidic networks for biochemical applications.     

Modulating patterns of two-phase flow with electric fields

This paper describes the use of electro-hydrodynamic actuation to control the transition between three major flow patterns of an aqueous-oil Newtonian flow in a microchannel: droplets, beads-on-a-string (BOAS), and multi-stream laminar flow. We observed interesting transitional flow patterns between droplets and BOAS as the electric field was modulated. The ability to control flow patterns of a two-phase fluid in a microchannel adds to the microfluidic tool box and improves our understanding of this interesting fluid behavior.Microfluidic technologies have found use in a wide range of applications, from chemical synthesis to biological analysis to materials and energy technologies.^1,2 In recent years, there has been increasing interest in two-phase flow and droplet microfluidics, owing to their potential for providing a high-throughput platform for carrying out chemical and biological analysis and manipulations.^3–8 Although droplets may be generated in many different ways, such as with electric fields or extrusion through a small nozzle,^9–12 the most common microfluidic methods are based on the use of either T-junctions or flow-focusing geometries with which uniform droplets can be formed at high frequency in a steady-state fashion.^13,14 Various operations, such as cell encapsulation, droplet fusion, splitting, mixing, and sorting, have also been developed, and these systems have been demonstrated for a wide range of applications, including cell analysis, protein crystallization, and material synthesis.^1–17In addition to forming discrete droplets, where a disperse phase is completely surrounded by a continuous phase, it is also possible in certain situations to have different phases flow side-by-side. In fact, multi-stream laminar flow, either of the same phase or different phases, has been exploited for both biochemical analysis and microfabrication.^1,2,18–20 Beads-on-a-string (BOAS) is another potential flow pattern, which has been attracting attentions in microfluidics field. BOAS flow, owing to its special flow structures, may be particularly useful in some applications, such as optical-sensor fabrication.²¹ In BOAS flow, queues of droplets are connected by a series of liquid threads, which makes them look like a fluid necklace with regular periods.^21–25 The BOAS pattern is easily found in nature, such as silk beads and cellular protoplasm, and is often encountered in industrial processes as well, such as in electrospinning and anti-misting.^21,22 In general, it is thought that BOAS structure occurs mostly in viscoelastic fluids²² and is an unstable structure, which evolves continually and breaks eventually.^21–29Flow patterns determine the inter-relations of fluids in a microdevice and are an important parameter to control. Common methods for adjusting microfluidic flow patterns include varying the fluid flow rates, fluid properties, and channel geometries. Additionally, the application of an electric field can be a useful supplement for adjusting microfluidic flow patterns, although most work in this area has been focused on droplets and in some cases also on multi-stream laminar flows.^30–33 Here, in addition to forming droplets and two-phase laminar flow with electro-hydrodynamic actuation, we also observed a new stable flow pattern in a non-viscoelastic fluid, BOAS flow. Such flow patterns may find use in controlling the interactions between droplets, such as limited mixing by diffusion between neighboring droplets.To generate droplets, we used the flow-focusing geometry (Figure 1(a)), in which aqueous phase (water) was flown down the middle channel and droplets were pinched off by the oil phase (1-octanol) from the two side channels at the junction; Figure 1(b) shows the droplets formed after the junction. To apply electric field along the main channel where the droplets were formed, we patterned a pair of electrodes upstream and downstream of the junction (Figure 1(a); for experimental details, please see Ref. ³⁴ for supplementary material). The average electric field strength may be calculated from the voltages applied and the distance (1.7 mm) between the two electrodes. When a high voltage was applied along the channel between the two electrodes, the aqueous-oil interface at the flow-focusing junction became charged and behaved like a capacitor. As a result, more negative charges were drawn back upstream towards the positive electrode, and left behind more positive charges at the aqueous-oil interface, which then became encapsulated into the aqueous droplets dispersed in the oil phase.Open in a separate windowFIG. 1.(a) Schematic of the setup. (b) Micrograph showing droplet generation in a flow-focusing junction. The scale bar represents 40 μm.The positively charged aqueous-oil interface was stretched under an applied electric field, and by adjusting the voltage and/or the two-phase flow-rate ratio, we found interestingly that various flow patterns emerged. We tested different combinations of applied voltages and flow-rate ratios and found that most of them resulted in similar flow patterns and transitions between flow patterns.Figure ​Figure22 illustrates the effects of varying the applied voltages on droplets at a fixed liquid flow rate. With increasing electric-field strength and force, we found it was easier for the aqueous phase to overcome interfacial tension and form droplets. For example, as the voltage increased from 0.0 kV to 0.8 kV (average field strength increased from 0 to 0.47 V/μm), droplet-generation frequencies became slightly higher, and the formed droplets were smaller in volume. Additionally, droplets gradually became more spherical in shape at higher voltages.Open in a separate windowFIG. 2.Images showing the effects of applied voltage on droplet shape and flow pattern. Oil-phase flow rate, 0.5 μl/min; aqueous-phase flow rate, 0.2 μl/min. The scale bar represents 40 μm.As the voltage increased further (e.g., up to 1.0 kV in Figure ​Figure3),3), the distance between neighboring droplets became smaller, and the aqueous-oil interface at the junction was stretched further toward the downstream channel. At a threshold voltage (1 kV here with corresponding average field strength of 0.59 V/μm), the tip of the aqueous-oil interface would catch up with the droplet that just formed, and the tip of the interface of this newly captured droplet would in turn catch up with the interface of the droplet that formed before it. Consequently, a series of threads would connect all the droplets flowing between the two electrodes, thus resulting in a BOAS flow pattern.Open in a separate windowFIG. 3.Series of images showing the reversibility and synchronicity of a transitional flow pattern between droplets and BOAS (bead-on-a-string). Voltage applied, 1.00 kV (corresponding field strength of 0.59 V/μm); oil-phase flow rate, 0.5 μl/min; aqueous-phase flow rate, 0.2 μl/min. The scale bar represents 40 μm.At voltages near the threshold value, the flow pattern was not stable, but oscillated between droplets flow and BOAS flow. Figure ​Figure33 is a series of images captured by a high-speed camera that show the flow in this transition region. In Figures 3(a) and 3(b), the string of BOAS became thinner over time, and then the BOAS broke into droplets (Figures 3(c) and 3(d)). The newly formed droplets, however, were not stable either. Thin liquid threads would appear and then connect neighboring droplets, and a new switching period between discrete droplets and BOAS would repeat (Figures 3(e)–3(h)). In addition to this oscillation and reversibility, the flow pattern had a synchronous behavior: all the droplets appeared connected simultaneously by liquid threads or were separated at the same time.When the voltage reached 1.3 kV, which corresponded to an average field strength of 0.76 V/μm, a stable BOAS flow was obtained (Figure 4(a)). BOAS structures are thought to be present mostly in viscoelastic fluids,²² because viscoelasticity is helpful in enhancing the growth of beads and in delaying breakup of the string; thus, the viscoelastic filament has much longer life time than its Newtonian counterpart. Here, with the help of electric field, regular BOAS structures are realized in a non-viscoelastic fluid (water) in microchannels.Open in a separate windowFIG. 4.(a) Micrograph showing BOAS flow in a channel. (b) Profile of the top-half of the BOAS flow recorded continuously at a cross-section (shown in Figure 4(a)) of a channel. Voltage applied, 1.30 kV (corresponding field strength of 0.76 V/μm); oil-phase flow rate, 0.5 μl/min; aqueous-phase flow rate, 0.2 μl/min. The scale bar represents 40 μm.Microenvironment and electric fields alter the common evolution of BOAS structure observed in macroscopic or unbound environments. The BOAS structure formed in our experiments is not a stationary pattern, but a steady-state flowing one. Electric-field force prevents liquid strings from breaking between beads, and thus plays a similar role as elastic force in viscoelastic fluids. Figure 4(b) shows the dynamic BOAS profile, obtained at a fixed plane (shown in Figure 4(a)) perpendicularly across the channel as the BOAS structure passed through it. Droplets and liquid-thread diameters were nearly constant during the sampling time. The longer term experiments (over 3 min) showed there were slight variations of the two diameters in time, but the essential BOAS structure still remained qualitatively the same as a whole.When the voltage was further increased, the string diameter became larger and the droplet diameter became smaller. Because of the low flow-rate ratio (0.4) between the aqueous phase and oil phase used in the experiment depicted in Figure ​Figure4,4, the flow did not further develop into a multi-stream laminar flow, as would be expected at a higher voltage, and instead became unstable and irregular. When the flow-rate ratio was increased to 1.0 and the voltage was adjusted to 3.0 kV (corresponding field strength of 1.76 V/μm), we observed a stable multi-stream laminar flow (Figure ​(Figure5).5). The aqueous stream flowed in the channel center surrounded by the oil phase on the sides. This experiment showed that higher electric-field strengths alone would not give rise to another stable flow pattern (i.e., multi-stream laminar flow), but a suitable flow-rate ratio of aqueous phase to oil phase is required for the formation of stable two-phase laminar flow.Open in a separate windowFIG. 5.Micrograph showing multi-stream two-phase laminar flow in the channel. Voltage applied, 3.00 kV (corresponding field strength of 1.76 V/μm); oil-phase flow rate, 0.5 μl/min; aqueous-phase flow rate, 0.5 μl/min. The scale bar represents 40 μm.The flow patterns we observed may be described by a phase diagram (Figure ​(Figure6),6), which depends on two dimensionless numbers: capillary number, Ca = μ_aU_a/σ, and electric Bond number, Bo_e = E²(εD/σ). Ca and Bo_e describe the ratio of viscous force to interfacial tension force and the ratio of electric-field force to interfacial tension force, respectively. Here, μ_a (1 mPa s), σ (8.5 mN/m), ε (7.1 × 10⁻¹⁰ F/m), E, U_a, and D are, respectively, the aqueous-phase viscosity, aqueous-oil interfacial tension, aqueous-phase permittivity, electric field strength, aqueous-phase velocity, and the hydraulic diameter of the channel at the junction. Figure ​Figure66 shows clearly that at higher Ca, flow pattern changes gradually from droplet to BOAS and to multi-stream laminar flow with increasing Bo_e, which indicates the increasing importance of the electric-field force compared with the interfacial tension force. At lower Ca, flow pattern and transition show similar trend with increasing Bo_e as in the higher Ca case, except that multi-stream laminar flow is not observed. The relatively higher viscous force at higher Ca may be necessary for transitioning to the multi-stream laminar flow regime. In addition, Figure ​Figure66 shows that the BOAS window at the lower Ca is smaller than that at the higher Ca.Open in a separate windowFIG. 6.Phase diagram showing different flow patterns in the Ca and Bo_e space. Hollow symbols: oil-phase flow rate, 0.5 μl/min; aqueous-phase flow rate, 0.5 μl/min. Solid symbols: oil-phase flow rate, 0.5 μl/min; aqueous-phase flow rate, 0.2 μl/min.In summary, we showed the ability to use electric fields to generate and control different flow patterns in two-phase flow. With the aid of an applied field, we were able to generate BOAS flow patterns in a non-viscoelastic fluid, a pattern that typically requires a viscoelastic fluid. The BOAS structure was stable and remained as long as the applied electric field was on. We also report transitional flow patterns, those between droplets and BOAS exhibited both good reversibility as well as synchronicity. And with a suitable flow-rate ratio between the two phases, BOAS flow could be transitioned into a stable two-phase laminar flow by applying a sufficiently high field strength. Finally, a phase diagram was presented to describe quantitatively the flow-pattern regimes using capillary number and electric Bond number. The phenomena we report here on the properties of two-phase flow under an applied electric field may find use in developing a different approach to exert control over droplet based or multi-phase laminar-flow based operations and assays, and also aid in understanding the physics of multi-phase flow.     

Plasmonic hotspots of dynamically assembled nanoparticles in nanocapillaries: Towards a micro ribonucleic acid profiling platform

Plasmonic hot spots, generated by controlled 20-nm Au nanoparticle (NP) assembly, are shown to suppress fluorescent quenching effects of metal NPs, such that hair-pin FRET (Fluorescence resonance energy transfer) probes can achieve label-free ultra-sensitive quantification. The micron-sized assembly is a result of intense induced NP dipoles by focused electric fields through conic nanocapillaries. The efficient NP aggregate antenna and the voltage-tunable NP spacing for optimizing hot spot intensity endow ultra-sensitivity and large dynamic range (fM to pM). The large shear forces during assembly allow high selectivity (2-mismatch discrimination) and rapid detection (15 min) for a DNA mimic of microRNA.Irregular expressions of a panel of microRNAs (miRNA) in blood and other physiological fluids may allow early diagnosis of many diseases, including cancer and cardiovascular diseases.¹ However, quantifying all relevant miRNAs (out of 1000), with similar sequences over 22 bases² and large variations in expression level (as much as 100 fold) at small copy numbers, requires a new molecular diagnostic platform with high-sensitivity, high-selectivity, and large dynamic range. Current techniques for miRNA profiling, such as Northern blotting,³ microarray-based hybridization,⁴ and real-time quantitative polymerase chain reaction⁵ are expensive and complex. A simple and rapid miRNA array would allow broad distribution of molecular diagnostic devices for cancer and chronic diseases, eventually into homes for frequent prescreening of many diseases.At their low concentrations in untreated samples, optical sensing of miRNA is most promising. Plasmonically excited Raman scattering (SERS) and fluorescence sensors from metallic nanoparticles (NPs) or surfaces have enhanced the sensitivity of optical molecular sensors by orders of magnitude.^{6, 7, 8, 9} However, probe-less SERS sensing or fluorescent sensing of unlabeled targets are insufficiently specific for miRNA targets in heterogeneous samples. Plasmonic detection is also very compatible with FRET probes whose donor dye offers small light sources to excite fluorescently labelled targets upon hybridization.^{7, 10}A particular family of FRET reporters does offer label-free sensing: hairpin oligo probes whose end-tagged fluorophores are quenched by the Au NP to which they are functionalized.¹¹ The fluorescent signal is only detected when the hairpin is broken by the hybridizing target nucleic acid or protein (for an aptamer probe), and the more rigid paired segment separates the end fluorophore from the quenching surface to produce a fluorescent signal. It is often hoped that plasmonics on the metal surface will enhance the intensity to overcome the quenching effect, if the linearized hairpin is within the NP plasmonic penetration length. However, since fluorescent quenching decays slowly (linearly) with fluorophore-metal spacing¹⁰ whereas the plasmonic intensity decays exponentially from a flat surface, careful experimentation shows that quenching dominates and the hairpin probe actually produces a larger intensity on non-metallic surfaces,¹⁰ on which it can not function as a label-free probe. Hence, only μM limit-of-detection (LOD) has been achieved with this technique on single NPs or on flat metal surfaces,¹² with expensive laser excitation and confocal detection.Plamonic hot spots formed between metal nanostructures and sharp nanocones can further amplify the plasmonic field.^{13, 14} The hot spot intensity decays algebraically with respect to the separation or cone tip distance and hence should dominate the linear decay of the metal quenching effect at some optimum separation.¹⁵ It is hence possible that plasmonic hot spots may allow much lower LOD with inexpensive optical instruments—ideally light-emitting diode light source and miniature camera. However, the dimension of the gaps, cones, and wedges needs to be at nanoscale, and the cost is now transferred to fabrication of such hot-spot substrates like bow-ties, double crescents, bull-eyes, etc.¹⁶ Low-cost wet-etching techniques for addressable nanocones that sustain converging plasmonic hot spots¹⁷ have been reported but the fabricated nanocones are often too non-uniform to allow precise quantification. NP monolayers have been shown to exhibit plasmonic hot spots and fluorescence enhancement.^{18, 19} However, the enhancement only occurs within a range of spacing between aggregated NPs, which is difficult to control and the location or even the existence of the hotspots are not known a priori.Higher sensitivity is expected if a minimum number of NPs are used in an assembly at a known location and if the NP assembly can produce crystal-like aggregates with controllable NP spacing. Induced DC and AC NP dipoles (related to dielectrophoresis) have been used to assemble NP crystals by embedded micro-electrodes to provide the requisite high field.^{20, 21} The resulting NP crystals are ideal for plasmonic hot spots, since the spacing of the regimented NP crystal can be controlled by the applied voltage. Conic nanocapillaries^{22, 23} will be used here for such field-induced NP assembly because the submicron-tip can focus the electric field into sufficient high intensity for NP assembly without embedded-electrodes. Because the field is highest at the tip due to field focusing, the micron-sized crystal would be confined to a small volume, which will be shown to be less than typical confocal volumes, at a known location. So long as the hotspots are regimented, the quantification of target molecules is determined by the total fluorescent intensity and is hence insensitive to the exact geometry of the nanocapillary.Fluorescent microscope equipped with tungsten lamp light source and normal CCD camera from Q Imaging were used for simultaneous optical and ion current measurements, as shown in Fig. ​Fig.1a.1a. The nanocapillaries were pulled from commercial glass capillaries using laser-assisted capillary puller. SEM image of a typical pulled glass nanocapillary in Fig. ​Fig.1b1b shows an inner diameter of 111 nm and cone angle of 7.3°. The capillary was inserted into a Polydimethylsiloxane chip with two reservoirs. The 20 nm Au NPs, functionalized with fluorescently labelled dsDNA, were injected into the base reservoir. With SEM imaging (Fig. S3 in the supplementary material²⁴), the functionalized DNA is found to prevent NP aggregation even in high ionic-strength Phosphate buffered saline buffer. The NP solution is then driven into the capillary through the tip by applying a positive voltage. Fig. ​Fig.1c1c shows the ion current evolution over 2 h at +1 V packing voltage. The ion current increases rapidly in the first 10 min, then at a much slower rate. The rise of current indicates assembly of conductive Au NP assembly at the tip. This was confirmed by the strong fluorescence signal at the tip region during the packing process (inset of Fig. ​Fig.1c).1c). The one-micron region (corresponding to roughly an aggregate volume of one attoliter) near the capillary tip shows a fluorescence signal after 1 min and also appeared as a dark spot in the transmission image (supplementary material, Fig. S1²⁴). This spot darkens with longer packing time but does not grow in size, consistent with the monotonically increasing ion current with increased packing density of the NP assembly. As contrast, a strong fluorescence appeared after only 1 min of packing, but the signal became weaker after 15 min (supplementary material, Fig. S1²⁴). This reduction in fluorescence is not due to bleaching of fluorophores because we took 2 images in 15 min at 5 s exposure each and control experiments show significant bleaching only beyond an exposure time of 100 s (see supplementary material).²⁴ Instead, the non-monotonic dependence of the fluorescence intensity with respect to time is because of the optimal hotspot spacing for highest plasmonic intensity at about 5–20 nm,^{25, 26, 27} which is reached at about 10 min.Open in a separate windowFigure 1Plasmonic hotspots generated at the tip of a nano-capillary. (a) Schematic of the experimental set up. (b) SEM image of glass nanocapillary shows opening at the tip with a diameter of 111 nm. (c) Current evolution during packing of fluorescently labeled gold particles at +1 V. Inset shows strong fluorescence only after 1 min of packing.The FRET probe is designed to exploit the plasmonic hotspot.²⁴ We first electrophoretically drove the target molecules in the tip side reservoir into the nano-capillary by applying a negative voltage of −1 V. During this process, the targets are trapped within the capillary and hybridize with the hairpin probes on the Au NP in the nanocapillary. Fluorescence of the unquenched hybridized probes is too weak to be detected by our detector as shown in Fig. ​Fig.2b.2b. A reverse positive voltage of +1 V was then applied to the capillary to pack the Au NPs to the tip. Due to plasmonic hot spots of aggregated gold nanoparticles, the fluorescence signal is significantly enhanced at the tip and can be detected by our CCD camera, as shown in Fig. ​Fig.2c2c.Open in a separate windowFigure 2(a) Schematics of designed hairpin probe on gold particle. (b) Before packing gold particles, probe fluorescence signal was too weak to be detect. (c) After packing for 3 minutes, a strong fluorescence signal appears at the NP aggregate. (d) Normalized intensity (average of all pixels above a threshold (15 au) normalized with respect to the average over all pixels (with 0-250 au)) as a function of packing voltage for different samples. Black, 1 nM target ; blue, 10 pM target; purple, 10 nM 2-mismatch non-target. (e) Intensity dependence on target concentration. Measured normalized intensity before packing (black) and after packing (red), for three independent experiments with different nano-capillaries at each concentration. NT stands for non-target at 10 nM as a reference.For the same packing time, the fluorescence intensity increases initially but saturates after 10 min time of trapping (supplementary material, Fig. S2(a)²⁴). Over 10 min of trapping with a negative voltage, we found the fluorescence intensity exhibits a maximum at a packing time of 3 min (supplementary material, Fig. S2(b)²⁴). In later experiments, we used 10 min trapping time and 3 min packing time as standards.Fig. ​Fig.2d2d shows the fluorescence intensity is sensitive to the positive packing voltage at different concentration of target and non-target molecules. For target samples (1 nM and 10pM), the optimal voltage is about 1 V. We suspect that with larger voltage, the NPs are packed too tightly such that the NP spacing is smaller than the optimal distance for plasmonic hotspots. The fluorescence intensity for a nontarget with two mismatches is 7 times lower than the target even with a 10 times higher concentration (10 nM). Moreover, the optimal voltage for the non-target miRNA is reduced to 0.5 V instead 1 V for the target miRNA. Strong shear during electrophoretic packing has probably endowed this high selectivity.²⁰Using the protocol above, the LOD and dynamic range of the target was determined (Fig. ​(Fig.2e).2e). The intensity at each concentration is measured with three independent experiments with different nanocapillaries to verify insensitivity with respect to the nanocapillary. The intensity increases monotonically with respect to the concentration from 1fM to 1pM. Beyond 1pM, the fluorescence signal saturates, presumably because all hairpin probes at the tip have been hybridized. At 1 fM, the fluorescent intensity is still well above the background measured from the non-target sample. Note both auto-fluorescence of gold nanoparticles and free diffusing non-target DNA molecules contribute to the background. Given the volume of tip side reservoir (∼50 μl), there are about 30 000 target molecules in the reservoir at 1 fM. However, with a short 10 min trapping time, we estimate only a small fraction of these molecules, less than 100, have been transferred from the tip reservoir into the nanocapillary.     

Effect of diabetic serum factor on polymorphonuclear leucocyte membrane hydrophobicity and particle internalisation

Membrane hydrophobicity and slalidase activity of normal Poly morphonuclear Leucocyte were significantly enhanced when incubated with DSF. As a consequence, internalisation ofE. coli andS. aureus (opsonised or unopsonised) were greatly dimnished, internalisation ofE. coli being higher in either category. Although, increase in hydrophobicity of the membrane correlated well with the time of decrease of particle internalisation (both at 30 min.), enhancement of sialidase activity did not coincide with the said alterations.     

Enhancing conjugation rate of antibodies to carboxylates: Numerical modeling of conjugation kinetics in microfluidic channels and characterization of chemical over-exposure in conventional protocols by quartz crystal microbalance

This research reports an improved conjugation process for immobilization of antibodies on carboxyl ended self-assembled monolayers (SAMs). The kinetics of antibody/SAM binding in microfluidic heterogeneous immunoassays has been studied through numerical simulation and experiments. Through numerical simulations, the mass transport of reacting species, namely, antibodies and crosslinking reagent, is related to the available surface concentration of carboxyl ended SAMs in a microchannel. In the bulk flow, the mass transport equation (diffusion and convection) is coupled to the surface reaction between the antibodies and SAM. The model developed is employed to study the effect of the flow rate, conjugating reagents concentration, and height of the microchannel. Dimensionless groups, such as the Damköhler number, are used to compare the reaction and fluidic phenomena present and justify the kinetic trends observed. Based on the model predictions, the conventional conjugation protocol is modified to increase the yield of conjugation reaction. A quartz crystal microbalance device is implemented to examine the resulting surface density of antibodies. As a result, an increase in surface density from 321 ng/cm², in the conventional protocol, to 617 ng/cm² in the modified protocol is observed, which is quite promising for (bio-) sensing applications.Microfluidics have been implemented in various bio-medical diagnostic applications, such as immunosensors and molecular diagnostic devices.¹ In the last decade, a vast number of biochemical species has been detected by microfluidic-based immunosensors. Immunosensors are sensitive transducers which translate the antibody-antigen reaction to physical signals. The detection in an immunosensor is performed through immobilization of an antibody that is specific to the analyte of interest.² The antibody is often bound to the transducing surface of the sensor covered by self-assembled monolayers (SAMs). SAMs are organic materials that form a thin, packed and robust interface on the surface of noble metals like that of gold, suitable for biosensing applications.³ Thiolic SAMs have a “head” group that shows a high affinity to being chemisorbed onto a substrate, typically gold. The SAMs'' carboxylic functional group of the “tail” end can be linked to an amine terminal of an antibody to form a SAM/antibody conjugation.^3,4 The conjugation process is usually accomplished in the presence of carbodiimides, such as 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC). A yield increasing additive, N-Hydroxysuccinimide (NHS), is often used to enhance the surface loading density of the antibody.^4,5A typical reaction for coupling the carboxylic acid groups of SAMs with the amine residue of antibodies in the presence of EDC/NHS is depicted in Figure ​Figure11.⁴ NHS promotes the generation of an active NHS ester (k₂ reaction path). The NHS ester is capable of efficient acylation of amines, including antibodies (k₃ reaction path). As a result, the amide bond formation reaction, which typically does not progress efficiently, can be enhanced using NHS as a catalyst.⁴Open in a separate windowFIG. 1.NHS catalyzed conjugation of antibodies to carboxylic-acid ended SAMs through EDC mediation (Adapted from G. T. Hermanson, Bioconjugate Techniques, 2nd. Edition. Copyright 2008 by Elsevier⁴). EDC reacts with the carboxylic acid and forms o-acylisourea, a highly reactive chemical that reacts with NHS and forms an NHS ester, which quickly reacts with an amine (i.e., antibody) to form an amide.A number of groups have studied EDC/NHS mediated conjugation reactions such as the ones depicted in Figure ​Figure1.1. The general stoichiometry of the reaction involves a carboxylic acid (SAM), an amine (antibody), and EDC to produce the final amide (antibody conjugated SAM) and urea. However, the recommended concentration ratio of the crosslinking reagents inside the buffer, i.e., the ratio of EDC and NHS with respect to adsorbates and each other, varies from one study to another.⁶ The kinetics of the reactions outlined in Figure ​Figure11 have also been investigated,^4,6–8 but only in the absence of NHS for EDC or carboxylic acids in aqueous solutions.⁸ A relatively recent experimental study verified the catalytic role of the yield-increasing reagent N-hydroxybenzotriazole (HOBt), which acts similarly to NHS.⁷ In this study, the amide formation rate (k₃ reaction path, Figure ​Figure1)1) was found to be dependent on the concentration of the carboxylic acid and EDC in the buffer solution, and independent of the amine and catalyst reagent concentration. The same group also showed that the amide bond formation kinetics is controlled by the reaction between the carboxylic acid and the EDC to give the O-acylisourea, which they marked as the rate-determining step (k₁ reaction path, Figure ​Figure11).The k₁ reaction path, or the conjugation reaction, is usually a lengthy process and takes between 1 and 3 h.^4,9 Compared to k₁, the k₂ and ?k₃ reactions are considerably faster. Microfluidics has the potential to enhance the kinetics of these reactions using the flow-through mode.^10,11 This improvement occurs because while conventional methods rely only on diffusion as the primary reagent transport mode, microfluidics adds convection to better replenish the reagents to the reaction surfaces. However, there are many fundamental fluidic and geometrical parameters that might affect the process time and reagents consumption in a microfluidics environment, such as concentration of antibodies and reagents, flow rate, channel height, and final surface density of antibodies. A model that studies the kinetics of conjugation reaction against all these parameters would therefore be helpful for the optimization of this enhanced kinetics.There are a number of reports on numerical examination of the kinetics of binding reactions in microfluidic immunoassays.^12–15 All these models developed so far couple the transport of reagents, by diffusion and convection, to the binding on the reaction surface. Myszka''s model assumes a spatially homogeneous constant concentration of reagents throughout the reaction chamber, thus fails to describe highly transport-limited conditions due to the presence of spatial heterogeneity and depletion of the bulk fluid from reagents.^16,17 In transport-limited conditions, the strength of reaction is superior to the rate of transport of reagents to the reaction surface.^18,19 More recently, the convection effects were included in a number of studies, describing the whole kinetic spectrum from reaction-limited conditions to transport-limited reactions.^20–22 Immunoreaction kinetics has also been examined with a variety of fluid driving forces, from capillary-driven flows,²⁰ to electrokinetic flows in micro-reaction patches,²¹ pressure-driven flows in a variety of geometric designs.²² Despite these comprehensive numerical investigations, the fundamental interrelations between the constitutive kinetic parameters, such as concentration, flow velocity, microchannel height, and antibody loading density, have not been studied in detail. In addition, the conjugation kinetics has not yet been exclusively examined.In this paper, a previous model for immunoreaction is modified to study the antibody/SAM conjugation reaction in a microfluidic system. Model findings are used to examine the process times recommended in the literature and possible modification scenarios are proposed. The new model connects the convective and diffusive transport of reagents in the bulk fluid to their surface reaction. The conjugation reaction is studied against fluidic and geometrical parameters such as flow rate, concentration, microchannel height and surface density of antibodies. Damköhler number is used to compare the reaction and fluidic phenomena present and justify the kinetic trends observed. Model predictions are discussed and the main finding on possible overexposure of carboxylates to crosslinking reagents, in conventional protocols, is verified by comparing the resultant antibody loading densities obtained using a quartz crystal microbalance (QCM) set up. The results demonstrate an improved receptor (antibody) loading density which is quite promising for a number of (bio-) sensing applications.^23,24 Major application areas include antibody-based sensors for on-site, rapid, and sensitive analysis of pathogens such as Bacillus anthracis,²³ Escherichia coli, and Listeria monocytogenes, and toxins such as fungal pathogens, viruses, mycotoxins, marine toxins, and parasites.²⁴     

Out of character: on the creation of virtuous machines

The emerging discipline of Machine Ethics is concerned with creating autonomous artificial moral agents that perform ethically significant actions out in the world. Recently, Wallach and Allen (Moral machines: teaching robots right from wrong, Oxford University Press, Oxford, 2009) and others have argued that a virtue-based moral framework is a promising tool for meeting this end. However, even if we could program autonomous machines to follow a virtue-based moral framework, there are certain pressing ethical issues that need to be taken into account, prior to the implementation and development stages. Here I examine whether the creation of virtuous autonomous machines is morally permitted by the central tenets of virtue ethics. It is argued that the creation of such machines violates certain tenets of virtue ethics, and hence that the creation and use of those machines is impermissible. One upshot of this is that, although virtue ethics may have a role to play in certain near-term Machine Ethics projects (e.g. designing systems that are sensitive to ethical considerations), machine ethicists need to look elsewhere for a moral framework to implement into their autonomous artificial moral agents, Wallach and Allen’s claims notwithstanding.     

An integrated microfluidic device for rapid serodiagnosis of amebiasis

A microfluidic device was successfully fabricated for the rapid serodiagnosis of amebiasis. A micro bead-based immunoassay was fabricated within integrated microfluidic chip to detect the antibody to Entamoeba histolytica in serum samples. In this assay, a recombinant fragment of C terminus of intermediate subunit of galactose and N-acetyl-_D-galactosamine-inhibitable lectin of Entamoeba histolytica (C-Igl, aa 603-1088) has been utilized instead of the crude antigen. This device was validated with serum samples from patients with amebiasis and showed great sensitivity. The serodiagnosis can be completed within 20 min with 2 μl sample consumption. The device can be applied for the rapid and cheap diagnosis of other infectious disease, especially for the developing countries with very limited medical facilities.Entamoeba histolytica is the causative agent of amebiasis and is globally considered a leading parasitic cause of human mortality.¹ It has been estimated that 50 × 10⁶ people develop invasive disease such as amebic dysentery and amebic liver abscess, resulting in 100 000 deaths per annum.^{2, 3} High sensitive diagnosis method for early stage amebiasis is quite critical to prevent and cure this disease. To date, various serological tests have been used for the immune diagnosis of amebiasis, such as the indirect fluorescent antibody test (IFA) and enzyme-linked immunosorbent assay (ELISA).We have recently identified a 150-kDa surface antigen of E. histolytica as an intermediate subunit (Igl) of galactose and N-acetyl-_D-galactosamine-inhibitable lectin.^{4, 5} In particular, it has been shown that the C-terminus of Igl (C-Igl, aa 603-1088) was an especially useful antigen for the serodiagnosis of amebiasis. ELISA using C-Igl is more specific than the traditional ELISA using crude antigen.⁶ However, the ELISA process usually takes several hours, which is still labor-intensive and requires experienced operators to perform. More economic and convenient filed diagnosis methods are still in need, especially for the developing countries with limited medical facilities.Among all the bioanalytical techniques, microfluidics has been attracting more and more attention because of its low reagent/power consumption, the rapid analysis speed as well as easy automation.^{7, 8, 9, 10, 11} Especially with the development of the fabrication technique, microfluidics chip can include valves, mixers, pumps, heating devices, and even micro sensors, so many traditional bioanalytical methods can be performed in the microfluidics. Qualitative and quantitative immune analysis on the microfluidic chip was successfully proved by plenty of research with improved sensitivity, shorten reaction time, and less sample consumption.^{8, 10, 11, 12, 13, 14, 15, 16, 17} Moreover, with the intervention of other physical, chemical, biology, and electronic technology, microfluidic technique has been successfully utilized in protein crystallization, protein and gene analysis, cell capture and culturing and analysis as well as in the rapid and quantitative detection of microbes.^{13, 14, 15, 16, 17, 18, 19, 20}Herein, we report a new integrated microfluidic device, which is capable of rapid serodiagnosis of amebiasis with little sample consumption. The microfluidic device was fabricated from polydimethysiloxane (PDMS) following standard soft lithography.^{21, 22} The device was composed of two layers (shown in Figure ​Figure1)1) including upper fluidic layer (in green and blue) and bottom control layer (in red).Open in a separate windowFigure 1Structure illustration of microfluidic chip.To create the fluidic layer and the control layer, two different molds with different patterns have fabricated by photolithographic processes. The mold to create the fluidic channels was made by positive photoresist (AZ-50 XT), while the control pneumatic mold was made by negative photoresist (SU8 2025). For the chip fabrication, the fluidic layer is made from PDMS (RTV 615 A: B in ratio 5:1), and the pattern was transferred from the respective mold. The control layer is made from PDMS (RTV 615 A:B in ratio 20:1). The two layers were assembled and bonded together accurately, and there is elastic PDMS membrane about 30 μm thick between the fluidic layer channels and control layer.^{21, 22} The elastic membrane at the intersection can deform to block the fluid inside the fluidic channels, functioning as valves under the pressures introduced though control channels. There are two types of channels in fluidic layer, the rectangular profiled (in green, 200 μm wide, 35 μm thick) channel and round profiled channels (in blue, 200 μm wide, 25 μm center height). Because of the position of the valves on the fluidic channels, two types of valves (Figure ​(Figure2a)2a) were built, working as a standard valve and a sieve valve. The standard valves (on blue fluidic channels) can totally block the fluid because of the round profile of fluidic channel; the sieve valve can only half close because of the rectangular profile. The sieve valve can be used to trap the microspheres (beads) filled inside the green fluidic channels, while letting the fluid pass through. By this sieve valve, a micro column (in green) is constructed, where the entire ELISA reaction happens. The micrograph of the fabricated micro device is shown in Figure ​Figure2b.2b. The channels were filled with food dyes in different colors to show the relative positions of the channels. The pressures though different control channels are individually controlled by solenoid valves, connected to a computer through relay board. By programming the status (on/off) of various valves at different time periods, all the microfluidic chip operation can be digitally controlled by the computer in manual, semi-automatic, or automatic manner.Open in a separate windowFigure 2(a) Structure illustration of micro column, standard valve and sieve valve; (b) photograph of the microfluidic chip.To validate this device, 12 patient serum samples were collected. Sera from 9 patients (Nos. 1–9) with an amebic liver abscess or amebic colitis were used as symptomatic cases. The diagnosis of these patients was based on their clinical symptoms, ultrasound examination (liver abscess) and endoscopic or microscopic examination (colitis). We also identified the clinical samples using PCR amplification of rRNA genes.²⁴ As negative control, sera obtained from 3 healthy individuals with no known history of amebiasis were mixed into pool sera. The serum was positive for E. histolytica with a titer of 1:64 (borderline positive), as determined by an indirect fluorescent-antibody (IFA) test.^{23, 24} In our previously study, the sensitivity and specificity of the recombinant C-Igl in the ELISA were 97% and 99%.^{6, 25} In the current study, the serodiagnosis of amebiasis was also examined by ELISA using C-Igl.²⁶ The cut-off for a positive result was defined as an ELISA value > 3 SD above the mean for healthy negative controls²⁷ (shown in Figure ​Figure3).3). The seropositivity to C-Igl was 100% in patients with amebiasis.Open in a separate windowFigure 3ELISA reactivity of sera from patients against C-Igl. ELISA plate was coated with 100 ng per well of C-Igl. Serum samples from patients and healthy controls were used at 1:400 dilutions. The dashed line indicates the cut-off value. Data are representative of results from three independent experiments.In the diagnosis process with microfluidic chip, the 4 micro immuno-columns filled with C-Igl-coated microspheres were the key components of the device. The C-Igl was prepared in E. coli as inclusion bodies. After expression, the recombinant protein was purified and analyzed by SDS-PAGE. The apparent molecular mass was 85 kDa.²⁶The immune-reaction mechanism is illustrated in Figure ​Figure4.4. The anti-His monocolonal antibody was immobilized onto the microspheres (beads, 9 μm diameter) coated with protein A. The C-Igl was then immobilized onto the beads through the binding between the His tag and C-Igl. For the diagnosis, the microspheres immobilized with C-Igl and blocked by 5% BSA were preloaded into the columns for the rapid analysis of the patient serum samples. Generally, serum samples which were diluted 100 times were first loaded into the reaction column and incubated at room temperature for 5 min. After being washed by PBS buffer, FITC-conjugated goat anti-human polyclonal antibody was added into the column for 4 min incubation. The fluorescence image can be collected by the fluorescence microscope after the micro column was washed with PBS buffer. From loading diluted serum samples into column to collecting fluorescence images, the total time to complete the immunoassay is less than 10 min. The final fluorescence results were analyzed by Image Pro Plus 6.0.Open in a separate windowFigure 4Schematic representation of the ELISA in the chip.Different reaction conditions have been investigated to find the optimized ones. For each patient, 2 μl sample is enough for the analysis. The designed microfluidic chip with 4 micro columns is capable for 4 parallel analyses at the same time. More micro columns can be integrated into the device if more parallel tests are needed.Different incubating time for the diagnosis has also been investigated and no significant difference has been found for various time periods. It is enough to incubate the chip for only 5 min. The total diagnosis time for one sample is less than 10 min. The detection result appeared as the fluorescence intensity of the reaction column. As shown in Figure ​Figure5,5, the negative sample showed relatively low fluorescence intensity, because little FITC-conjugated goat anti-human polyclonal antibody could attach to the surface of microspheres; on the contrast, the positive sample showed much brighter fluorescence. The fluorescence intensity can be transferred to digital data (Table

Water transport to the core–mantle boundary

Water is transported to Earth''s interior in lithospheric slabs at subduction zones. Shallow dehydration fuels hydrous island arc magmatism but some water is transported deeper in cool slab mantle. Further dehydration at ∼700 km may limit deeper transport but hydrated phases in slab crust have considerable capacity for transporting water to the core-mantle boundary. Quantifying how much remains the challenge.

Water can have remarkable effects when exposed to rocks at high pressures and temperatures. It can form new minerals with unique properties and often profoundly affects the physical, transport and rheological properties of nominally anhydrous mantle minerals. It has the ability to drastically reduce the melting point of mantle rocks to produce inviscid and reactive melts, often with extreme chemical flavors, and these melts can alter surrounding mantle with potential long-term geochemical consequences. At the base of the mantle, water can react with core iron to produce a super-oxidized and hydrated phase, FeO₂H_x, with the potential to profoundly alter the mantle and even the surface and atmosphere redox state, but only if enough water can reach such depths [1].Current estimates for bulk mantle water content based on the average H₂O/Ce ratio of oceanic basalts from melt inclusions and the most un-degassed basalts, coupled with mass balance constraints for Ce, indicate a fraction under one ocean mass [2], a robust estimate as long as the basalts sampled at the surface tap all mantle reservoirs. The mantle likely contains some primordial water but given that the post-accretion Earth was very hot, water has low solubility and readily degasses from magma at low pressures, and its solubility in crystallizing liquidus minerals is also very low, the mantle just after accretion may have been relatively dry. Thus, it is plausible that most or even all of the water in the current mantle is ‘recycled’, added primarily by subduction of hydrated lithospheric plates. If transport of water to the core–mantle boundary is an important geological process with planet-scale implications, then surface water incorporated into subducting slabs and transported to the core–mantle boundary may be a requirement.Water is added to the basaltic oceanic crust and peridotitic mantle in lithospheric plates (hereafter, slab crust and slab mantle, respectively) at mid-ocean ridges, at transform faults, and in bending faults formed at the outer rise prior to subduction [3]. Estimates vary but about 1 × 10¹² kg of water is currently subducted each year into the mantle [4], and at this rate roughly 2–3 ocean masses could have been added to the mantle since subduction began. However, much of this water is returned to the surface through hydrous magmatism at convergent margins, which itself is a response to slab dehydration in an initial, and large, release of water. Meta-basalt and meta-sediments comprising the slab crust lose their water very efficiently beneath the volcanic front because most slab crust geotherms cross mineral dehydration or melting reactions at depths of less than 150 km, and even if some water remains stored in minerals like lawsonite in cooler slabs, nearly complete dehydration is expected by ∼300 km [5].Peridotitic slab mantle may have much greater potential to deliver water deeper into the interior. As shown in Fig. 1a, an initial pulse of dehydration of slab mantle occurs at depths less than ∼200 km in warmer slabs, controlled primarily by breakdown of chlorite and antigorite when slab-therms cross a deep ‘trough’, sometimes referred to as a ‘choke point’, along the dehydration curve (Fig. 1a) [6]. But the slab mantle in cooler subduction zones can skirt beneath the dehydration reactions, and antigorite can transform directly to the hydrated alphabet silicate phases (Phases A, E, superhydrous B, D), delivering perhaps as much as 5 wt% water in locally hydrated regions (e.g. deep faults and fractures in the lithosphere) to transition zone depths [6]. Estimates based on mineral phase relations in the slab crust and the slab mantle coupled with subduction zone thermal models suggest that as much as 30% of subducted water may have been transported past the sub-volcanic dehydration front and into the deeper mantle [4], although this depends on the depth and extent of deep hydration of the slab mantle, which is poorly constrained. Coincidentally, this also amounts to about one ocean mass if water subduction rates have been roughly constant since subduction began, a figure tantalizingly close to the estimated mantle water content based on geochemical arguments [2]. But what is the likely fate of water in the slab mantle in the transition zone and beyond?Open in a separate windowFigure 1.(a) Schematic phase relations in meta-peridotite modified after [6,10,12]. Slab geotherms are after those in [4]. Cold slabs may transport as much as 5 wt% water past ‘choke point 1’ in locally hydrated regions of the slab mantle, whereas slab mantle is dehydrated in warmer slabs. Colder slab mantle that can transport water into the transition zone will undergo dehydration at ‘choke point 2’. How much water can be transported deeper into the mantle and potentially to the core depends on the dynamics of fluid/melt segregation in this region. (b) Schematic showing dehydration in the slab mantle at choke point 2. Migration of fluids within slab mantle will result in water dissolving into bridgmanite and other nominally anhydrous phases with a bulk storage capacity of ∼0.1 wt%, potentially accommodating much or all of the released water. Migration of fluids out of the slab into ambient mantle would also hydrate bridgmanite and other phases and result in net fluid loss from the slab. Conversely, migration of hydrous fluids into the crust could result in extensive hydration of meta-basalt with water accommodated first in nominally anhydrous phases like bridgmanite, Ca-perovskite and NAL phase, but especially in dense SiO₂ phases (stishovite and CaCl₂-type) that can host at least 3 wt% water (∼0.6 wt% in bulk crust).Lithospheric slabs are expected to slow down and deform in the transition zone due to the interplay among the many factors affecting buoyancy and plate rheology, potentially trapping slabs before they descend into the lower mantle [7]. If colder, water-bearing slabs heat up by as little as a few hundred degrees in the transition zone, hydrous phases in the slab mantle will break down to wadsleyite or ringwoodite-bearing assemblages, and a hydrous fluid (Fig. 1a). Wadselyite and ringwoodite can themselves accommodate significant amounts of water and so hydrated portions of the slab mantle would retain ∼1 wt% water. A hydrous ringwoodite inclusion in a sublithospheric diamond with ∼1.5 wt% H₂O may provide direct evidence for this process [8].But no matter if slabs heat up or not in the transition zone, as they penetrate into the lower mantle phase D, superhydrous phase B or ringwoodite in the slab mantle will dehydrate at ∼700–800 km due to another deep trough, or second ‘choke point’, transforming into an assemblage of nominally anhydrous minerals dominated by bridgmanite (∼75 wt%) with, relatively, a much lower bulk water storage capacity (< ∼0.1 wt%) [9] (Fig. 1a). Water released from the slab mantle should lead to melting at the top of the lower mantle [10], and indeed, low shear-wave velocity anomalies at ∼700–800 km below North America may be capturing such dehydration melting in real time [11].The fate of the hydrous fluids/melts released from the slab in the deep transition zone and shallow lower mantle determines how much water slabs can carry deeper into the lower mantle. Presumably water is released from regions of the slab mantle where it was originally deposited, like the fractures and faults that formed in the slab near the surface [3]. If hydrous melts can migrate into surrounding water-undersaturated peridotite within the slab, then water should dissolve into bridgmanite and coexisting nominally anhydrous phases (Ca-perovskite and ferropericlase) until they are saturated (Fig. 1b). And because bridgmanite (water capacity ∼0.1 wt%) dominates the phase assemblage, the slab mantle can potentially accommodate much or all of the released water depending on details of how the hydrous fluids migrate, react and disperse. If released water is simply re-dissolved into the slab mantle in this way then it could be transported deeper into the mantle mainly in bridgmanite, possibly to the core–mantle boundary. Water solubility in bridgmanite throughout the mantle pressure-temperature range is not known, so whether water would partially exsolve as the slab moves deeper stabilizing a melt or another hydrous phase, or remains stable in bridgmanite as a dispersed, minor component, remains to be discovered.Another possibility is that the hydrous fluids/melts produced at the second choke point in the slab mantle at ∼700 km migrate out of the slab mantle, perhaps along the pre-existing fractures and faults where bridgmanite-rich mantle should already be saturated, and into either oceanic crust or ambient mantle (Fig. 1b). If the hydrous melts move into ambient mantle, water would be consumed by water-undersaturated bridgmanite, leading to net loss of water from the slab to the upper part of the lower mantle, perhaps severely diminishing the slab’s capacity to transport water to the deeper mantle and core. But what if the water released from slab mantle migrates into the subducting, previously dehydrated, slab crust?Although slab crust is expected to be largely dehydrated in the upper mantle, changes in its mineralogy at higher pressures gives it the potential to host and carry significant quantities of water to the core–mantle boundary. Studies have identified a number of hydrous phases with CaCl₂-type structures, including δ-AlOOH, ϵ-FeOOH and MgSiO₂(OH)₂ (phase H), that can potentially stabilize in the slab crust in the transition zone or lower mantle. Indeed, these phases likely form extensive solid solutions such that an iron-bearing, alumina-rich, δ-H solid solution should stabilize at ∼50 GPa in the slab crust [12], but only after the nominally anhydrous phases in the crust, (aluminous bridgmanite, stishovite, Ca-perovskite and NAL phase) saturate in water. Once formed, the δ-H solid solution in the slab crust may remain stable all the way to the core mantle boundary if the slab temperature remains well below the mantle geotherm otherwise a hydrous melt may form instead [12] (Fig. 1a). But phase δ-H solid solution and the other potential hydrated oxide phases, intriguing as they are as potential hosts for water, may not be the likely primary host for water in slab crust. Recent studies suggest a new potential host for water—stishovite and post-stishovite dense SiO₂ phases [13,14].SiO₂ minerals make up about a fifth of the slab crust by weight in the transition zone and lower mantle [15] and recent experiments indicate that the dense SiO₂ phases, stishovite (rutile structure—very similar to CaCl₂ structure) and CaCl₂-type SiO₂, structures that are akin to phase H and other hydrated oxides, can host at least 3 wt% water, which is much more than previously considered. More importantly, these dense SiO₂ phases apparently remain stable and hydrated even at temperatures as high as the lower mantle geotherm, unlike other hydrous phases [13,14]. And as a major mineral in the slab crust, SiO₂ phases would have to saturate with water first before other hydrous phases, like δ-H solid solution, would stabilize. If the hydrous melts released from the slab mantle in the transition zone or lower mantle migrate into slab crust the water would dissolve into the undersaturated dense SiO₂ phase (Fig. 1b). Thus, hydrated dense SiO₂ phases are possibly the best candidate hosts for water transport in slab crust all the way to the core mantle boundary due to their high water storage capacity, high modal abundance and high-pressure-temperature stability.Once a slab makes it to the core–mantle boundary region, water held in the slab crust or the slab mantle may be released due to the high geothermal gradient. Heating of slabs at the core–mantle boundary, where temperatures may exceed 3000°C, may ultimately dehydrate SiO₂ phases in the slab crust or bridgmanite (or δ-H) in the slab mantle, with released water initiating melting in the mantle and/or reaction with the core to form hydrated iron metal and super oxides, phases that may potentially explain ultra-low seismic velocities in this region [1,10]. How much water can be released in this region from subducted lithosphere remains a question that is hard to quantify and depends on dynamic processes of dehydration and rehydration in the shallower mantle, specifically at the two ‘choke points’ in the slab mantle, processes that are as yet poorly understood. What is clear is that subducting slabs have the capacity to carry surface water all the way to the core in a number of phases, and possibly in a phase that has previously seemed quite unlikely, dense SiO₂.     

Quantifying the volume of single cells continuously using a microfluidic pressure-driven trap with media exchange

We demonstrate a microfluidic device capable of tracking the volume of individual cells by integrating an on-chip volume sensor with pressure-activated cell trapping capabilities. The device creates a dynamic trap by operating in feedback; a cell is periodically redirected back and forth through a microfluidic volume sensor (Coulter principle). Sieve valves are positioned on both ends of the sensing channel, creating a physical barrier which enables media to be quickly exchanged while keeping a cell firmly in place. The volume of individual Saccharomyces cerevisiae cells was tracked over entire growth cycles, and the ability to quickly exchange media was demonstrated.Measuring cell growth is of primary interest to researchers who seek to study the effects of drugs, nutrients, disease, and environmental stress. This has traditionally been accomplished by monitoring the optical transmittance of large ensembles of cells and applying the Beer-Lambert Law.^1,2 Such population-scale measurements provide important culture statistics, but averaging obscures the behaviour of individual cells. In addition, these techniques often require cell synchronicity in order to correlate growth with specific points in the cell cycle, but synchronicity typically decays rapidly in many cell lines including Saccharomyces cerevisiae (yeast) cultures.³ Researchers have thus adopted methods that study the growth of individual cells. Quantifying cellular growth is especially challenging since proliferating cells such as yeast or Escherichia coli are irregularly shaped, and will only increase in size by a factor of two.⁴ Growth will affect the mass, volume, and density of the cell; having access to each of these characteristics is important in obtaining a complete picture of this process. Time-lapse fluorescence microscopy can provide valuable information as to the cell cycle progression of individual cells,⁵ but 2D optics requires geometric assumptions, and, thus, can provide an incomplete picture of growth.^6,7Microfluidic lab-on-chip devices with integrated sensors can provide high-resolution growth tracking of individual cells, either through mass, volume, or density monitoring.^4,7,8 Recently, a microfluidic mass sensor was used to track the buoyant mass of individual cells using a suspended microchannel resonator (SMR).^4,9 Monitoring growth can also be accomplished by tracking volume using microfluidic volume sensors⁷ operating on the Coulter principle.¹⁰ Trapping can be achieved by either (1) cycling the target back and forth through the sensor (pressure-driven⁴ and electrokinetic⁷) or (2) holding a cell in place (posts,¹¹ chevron structure,¹² and E-Field¹³). The former, dynamic approach, allows a single cell to be sampled periodically by reversing flow directions after a cell is detected. Simple in its implementation, this technique also has the ability to compensate for a drifting baseline current resulting from parasitic ionic changes within the sensing channel or other sources of noise. On the other hand, static traps allow cells to be held in place while the buffer is rapidly exchanged.¹² The ability to dynamically change cellular growth conditions during an experiment can lead to significant insight into the behaviour of cells in environments of varying salinity,¹⁴ oxidative,^15,16 or osmotic conditions,¹⁷ as well as the effect of nutrients¹⁸ and drugs.¹⁹In this work, we propose a device capable of tracking growth using high-resolution volume measurements, combining the best attributes of both types of measurement systems; continuous baseline correction and the ability to rapidly exchange cell media. This is accomplished by using a pressure-driven, feedback-based dynamic trap, whereby a cell is cycled back and forth through the sensor within a microfluidic channel. On-chip sieve valves positioned at both ends of the sensing channel are able to selectively capture a cell while the solution is being replaced. As proof of principle, the volume of several individual yeast cells was monitored over the course of their respective growth cycles, and the ability to quantify growth response to media exchange was demonstrated.Devices were fabricated using multilayered soft lithography with polydimethylsiloxane (PDMS) molding.²⁰ The completed device is pictured in Figure 1(a); full fabrication protocols are presented as supplementary material.²¹ To maximize measurement sensitivity, it is optimal to choose a channel width and height slightly larger than the dimensions of the target cell.²² However, yeast cells are asymmetrically shaped and tend to tumble as they traverse the sensor. Preliminary testing suggested this effect could be mitigated by having cells flow along trajectories far from the electrodes (through buoyancy), where electric field is more uniform. Thus, a channel height of 20 μm was chosen as a compromise. Channel height increases to 28 μm in the wider part of the central and bypass channels, a result of using a mold made out of reflowed photoresist.²³ Channel width was set at 25 μm through the sensor, and widens to 80 μm at the sieve valves to facilitate valve actuation, which requires a high width to height ratio.²⁰ The fluidic layer is integrated in a 35 μm thick PDMS spin-coated layer, above which sits a 50 μm tall valve channel in a 4 mm PDMS layer. Tubing connects I1 and I2 to a common inlet vial, V1 and V2 to vials filled with deionised water and O1 and O2 connect to empty vials (not pictured). Inlet pressures I1 and I2, and valve pressures V1 and V2 are controlled with manual regulators (SMC IR2000-N02-R and SMC IR2010-N02-R); outlet pressures are computer-controlled (SMC ITV-1011). This pressure scheme is detailed elsewhere.²⁴ Current pulses caused by transiting particles/cells (Figure 1(d)) were acquired by applying a 50 kHz, 220 mV AC voltage between a pair of electrodes and measuring the drawn current. This frequency is sufficiently elevated to avoid the electrical double layer capacitance at the electrode-electrolyte interface,²⁵ but low enough to avoid sensitivity to cell impedance or substrate.²⁶ The electrical setup used for these experiments has been described previously.^24,27 A temperature controller maintains the device at 30 °C.Open in a separate windowFIG. 1.(a) Micrograph of the microfluidic device. Two parallel bypass channels are connected by a sensing channel with sensing electrodes. Pressure is applied at inlets (I1, I2) and outlets (O1, O2) to control flow conditions. Valves (V1, V2) are positioned over each end of the sensing channel. Food coloring is used to highlight the valve (red) and fluidic layers (blue). (b) Flow mode: valves are unpressurized, and cells flow freely through the device. (c) Trapping mode: valves are pressurized to capture a cell within the central channel. Pressure-driven flow cycles the cell back and forth across the sensor. (d)Typical current pulses measured for a yeast cell.The cell capture, media exchange, and detection process occurs as follows. A cell suspension is loaded into the bypass channel and made to flow through the central sensing channel by imposing a pressure gradient (Figure 1(b)). Cells flowing through the sensor are observed optically; once a cell of interest is observed (a cell without a bud), valves are sealed (V1 = V2 = 35 psi). This stops all flow through the sensor, and enables bypass channels to be flushed and replaced with fresh media. After 2 min, valve channels are pressurized to 24 psi where they compress the channel to a sufficient height to physically restrict the passage of yeast cells, while allowing the media to flow through the central channel (Figure 1(c)). The pressure gradient between bypasses causes the media in the central channel to be flushed out, while the target cell is physically trapped. Replacing the media in the central channel takes 2 min. At this stage, a pressure-driven feedback-based dynamic trap can be initiated. In this dynamic trap mode, the pressure settings at O1 and O2 are adjusted to redirect the cell back and forth through the sensor, based on current pulses measured from cells transiting through the sensor. Through custom LabView® software, these outlet pressure settings are feedback-adjusted to maintain a speed of 250 μm/s in both directions at a detection frequency of 30 cells/min (Figure 1(d)). To minimize the effects of channel stretching/shrinking, the sum of pressures at O1 and O2 is held constant. This precaution was taken since the sensing channel structured within the flexible PDMS polymer will alter its geometry based on internal pressure.²⁸ The short central channel ensures steady nutrient replenishment from the bypasses. For example, a glucose molecule takes ∼4 min to diffuse from the bypass to the electrodes. In practice, Taylor-Aris dispersion will reduce this replenishment time considerably. Based on video analysis, 25% of the central channel''s media is replenished every pressure reversal (video presented as supplementary material²¹). Polystyrene microspheres of 3.9 ± 0.3 μm, 5.6 ± 0.2 μm, and 8.3 ± 0.7 μm (NIST size standards) were used to calibrate the sensor, and obtain the current pulse-to-volume calibration for every solution (supplementary material²¹). The validity of this calibration method is discussed elsewhere.²⁹ Care was taken to limit trajectory-based variations in signal: the device is positioned with electrodes at the top of the sensing channel, and with the negatively buoyant cells/particles flowing along the bottom. Based on previous experimental and theory work, we found that signal amplitude can vary as much as 3.5 fold for different heights.²⁷ The effect of trajectory on current pulse amplitude has also been reported elsewhere.^30,31 In this work, buoyancy is used to ensure that the cell flows along a trajectory at the same distance from the electrodes for every measurement.Saccharomyces cerevisiae (BY4743 Mat a/alpha, genotype: his3Δ1/his3Δ1 leu2Δ0/leu2Δ0 LYS2/lys2Δ0 met15Δ0/MET15 ura3Δ0/ura3Δ0 ade2::LEU2/ade2::URA3) was cultured to exponential phase at 30 °C in an incubator/shaker in yeast bacto-peptone (YPD) with 2% w/v glucose, supplemented with 0.2 M NaCl, 0.05% bovine serum albumin (BSA) and 42 mg/l adenine. Sodium chloride was added to enable the current pulse measurement, at a concentration where cells are viable;³² BSA was used to prevent cell agglomeration; adenine was supplemented since this particular yeast mutant does not produce its own supply. A cell suspension was introduced into the device, from which a cell at the early stages of its cell cycle was captured, and dynamically trapped for 100 min. Three typical cell growth results are shown in Figure 2(a). Since the culture was not synchronized, this leads to variability between “initial” cell volumes: there is a 27% difference in initial volume between the cells identified by red squares and green triangles. This is caused by (1) optical limits, whereby cells chosen for study are not all at the exact same cell cycle stage and (2) differences in the age of the mother cell: the more buds a mother cell has produced, the larger it becomes.³³ On average, captured yeast cell demonstrated a doubling time consistent with growth rates under ideal incubator/shaker conditions; nutrient depletion, electric field, and shear stresses are not affecting growth. Optical inspection of budding cells confirms that most growth is occurring at the daughter cell, as expected.³³ An elevated signal-to-noise ratio allows for high resolution volumetric measurements (4 μm³); cell asymmetry⁷ and trajectory variability^27,30,31 lead to a relative standard deviation of 6% for cells and 4% for microspheres of similar size. While mass or protein synthesis methods have indicated linear³⁴ or exponential^4,6,35,36 growth curves, volume-based methods have suggested sigmoidal patterns.^7,37 Prior to daughter cell emergence, and later in the cycle as the daughter cell emerges, volumetric growth rate declines.³⁸ In this work, it is difficult to ascertain with mathematical rigor the shape of the growth profile; however, for each cell, volume increases steadily throughout the growth cycle before declining near the end of the cycle.Open in a separate windowFIG. 2.(a) Growth curves for 3 cells trapped in succession. Simultaneous optical and electrical measurements allow cell cycle stage to be correlated with volume. Pictures of cell corresponding to the red squares are presented in 15 min increments. A cell is cycled through the sensor every 2 s. For clarity, each data point for yeast volume represents the average of data points over a period of 5 min, with standard deviation. (b) Demonstration of an interrupted growth cycle, where YPD + 0.2 M NaCl was replaced with 0.2 M NaCl at 40 min, and then again returned to YPD + 0.2 M NaCl at 80 min. The media exchange process takes 4 min.To demonstrate our ability to easily exchange media while maintaining a trap, the solution was exchanged 40 min into a yeast growth cycle; culture media was replaced with a pure saline solution 0.2 M NaCl + 0.05% BSA, and then replaced again with culture media at 80 min (Figure 2(b)). Cell growth is halted temporarily while in saline solution, before resuming normal growth thereafter. The cell cycle time is extended by this period. The cell volume drifts downward after the initial solution change at 40 min. Though this drift lies within our uncertainty bounds, cellular responses to osmotic shock on similar timescales have been documented elsewhere.³⁹ This result demonstrates an ability to quickly exchange cell media, and observe cellular response.In conclusion, we have demonstrated a microfluidic device capable of maintaining a dynamic, pressure-driven cell trap, which can monitor cellular volume over the cell cycle. Concurrent optical microscopy allows for real-time visual inspection of the cells. In addition, sieve valve integration provides for the exchange of media or the addition of drugs. Such a platform could also be key in cancer cell cytotoxicity assays,⁴⁰ where growth response to anticancer drugs could be monitored.     

Mixed control of uncertain jumping time-delay systems

In this paper, we investigate a class of linear continuous-time systems with Markovian jump parameters. An integral part of the system dynamics is a delayed state with time-varying and bounded delays. The jumping parameters are modeled as a continuous-time, discrete-state Markov process. Employing norm-bounded parametric uncertainties and utilizing the second-method of Lyapunov, we examine the problem of designing a mixed controller which minimizes a quadratic performance measure while satisfying a prescribed -norm bound on the closed-loop system. It is established that sufficient conditions for the existence of the mixed controller and the associated performance upper bound could be cast in the form of linear matrix inequalities.     

自动驾驶汽车的“道德算法”困境

当前自动驾驶汽车发展所面临伦理困境的一个核心问题是：是否应该将道德规范嵌入算法结构以及应当以何种方式嵌入。在面对未来可能的交通事故时，屏蔽信息而依靠“道德运气”进行随机选择和基于完全信息的人工智能系统自主抉择都存在严重困难，因此应当为自动驾驶汽车预设“道德算法”。而对于如何决定“道德算法”的问题，鉴于现有道德原则间的相互冲突、道德决策的复杂性以及人类道德判断的情境化特点，基于某种人类既定的道德原则或道德规范是不现实的。