**2. Opto‐electro‐reconfigurable‐microchannel (OERM)**

The first issue with designing OERM are the processes of melting and freezing channels of micron size. The problem of phase transition belongs to the Stefan problem category. It involves a moving boundary thatrepresents the phase change location [6]. Such problems have been extensively studied [7]. Their resolution is often strenuous and requires numerical methods [8, 9]. It is beyond the scope of this chapter. Nevertheless, phase transition can still be investigated without the need of a complete solution. Phase transition mainly affects the time‐dependence of the temperature profiles. It has, however, an indirect impact on the steady state that stems from the difference between water and ice thermal conductivities. It creates a discontinuity in the temperature gradient. More precisely, the first law of thermodynamics imposes that the heat flux J(x, t) be continuous. The boundary conditions at the water/ice interface located in x0(t) with water and ice respectively on the left and right side of the interface.are then u(x0(t)) = 0 and*kwater* ∂ *u* <sup>∂</sup> *<sup>x</sup>* (*x*<sup>0</sup> ‐ (*t*)) <sup>=</sup>*kice* ∂ *u* <sup>∂</sup> *<sup>x</sup>* (*x*<sup>0</sup> +(*t*)). As a result, at steady state, the temperature profile in the ice region is:

for *x* ∈ *x*0*∞*, *L* :*u∞ tr*(*x*)=*u∞*(*x*)= *<sup>J</sup> kice* (*L* ‐ *x*) + *Tc*. Tc is the temperature at x = L and symbolizes the cooler's temperature.

If a heat flux J=1.2\*104 W/cm2 calculated to impose T = 0°C in x = 100μm is used, the interface will be at *x* = *x∞* = 100*μm*: the microchannel is still 100μm deep at steady state. Nonethe‐ less, the temperature gradient in the liquid region is higher since the heat conductivity is lower. The steady state solution for *x* ∈ *x*0*∞*, *L* is easily derived, and the complete steady state solution is:

$$\begin{aligned} \mu\_{\stackrel{\text{def}}{\circ}}(\infty) &= \frac{l}{k\_{\text{water}}} (\text{x}\_{0\circ\circ} - \text{x}) \text{ if } \mathfrak{x} \in \left[0, \ \text{x}\_{0\circ}\right] \\\\ \mu\_{\stackrel{\text{def}}{\circ}}(\infty) &= \frac{l}{k\_{\text{ice}}} (L - \infty) + \ \text{T}\_c \text{ if } \mathfrak{x} \in \left[\mathfrak{x}\_{0\circ\circ} \quad L\ \right] \end{aligned}$$

Essentially, the temperature gradient is higher in the liquid region. It leads to higher temper‐ atures than in the case where only ice is considered. For instance it leads to a steady state temperature in x = 0 of value *T* =*u∞ tr*(0)= 2.8 °*C* = 276*K*

*Transient State*. The influence of phase transition on the transient state is now addressed. The temperature and heat flux solution of the heat conduction problem are plotted (see Figure 1) to give more insight on the physics that take place. It can be seen that the temperature reaches the melting point after 0.8s. Up to that point a single phase solution to the heat conduction problem is accurate since no phase transition occurs. Figure 1 illustrates that the temperature and heat flux profile when melting begins are already close to steady state. In particular, heat flux is almost constant, which is detrimental to melting. An approximate calculation of the melting time can be performed based on those curves. Since the melting region is small, the problem can be linearized for a first‐order approximation. The heat used for melting the 100μm region then results from the difference between the heat flux entering the region and the heat flux leaving the region. The heat flux J entering at x = 0 is constant and has been calculated previously. The heat flux exiting the region at x = 100 *μ*m and t = 0.8s is inferred from the single phase heat conduction problem solution. Their difference equals 280W/m2 . With H the water enthalpy of fusion per unit volume, H = 3.32\*108 J.m‐<sup>3</sup> , and l the length of the melting zone, l = 100μm, the first‐order approximation for the melting time is:

$$\sigma\_{melting}^{'} = \frac{Hl}{f - f(x - 100\mu m\_r \ t - 0.8s)} = 119s \approx 2\,\text{min}$$

to be in 37o

300

Biomedical Engineering

C environment. Therefore, an additional microfluidic system operated at 37o

to be designed and integrated seamlessly with the OERM system for accomplishing the biological studies. In this article, we will introduce the design and manufacturing methods of

Micro-Nano Mechatronics — New Trends in Material, Measurement, Control, Manufacturing and Their Applications in

The first issue with designing OERM are the processes of melting and freezing channels of micron size. The problem of phase transition belongs to the Stefan problem category. It involves a moving boundary thatrepresents the phase change location [6]. Such problems have been extensively studied [7]. Their resolution is often strenuous and requires numerical methods [8, 9]. It is beyond the scope of this chapter. Nevertheless, phase transition can still be investigated without the need of a complete solution. Phase transition mainly affects the time‐dependence of the temperature profiles. It has, however, an indirect impact on the steady state that stems from the difference between water and ice thermal conductivities. It creates a discontinuity in the temperature gradient. More precisely, the first law of thermodynamics imposes that the heat flux J(x, t) be continuous. The boundary conditions at the water/ice interface located in x0(t) with water and ice respectively on the left and right side of the

‐ (*t*)) <sup>=</sup>*kice*

will be at *x* = *x∞* = 100*μm*: the microchannel is still 100μm deep at steady state. Nonethe‐ less, the temperature gradient in the liquid region is higher since the heat conductivity is lower. The steady state solution for *x* ∈ *x*0*∞*, *L* is easily derived, and the complete steady

Essentially, the temperature gradient is higher in the liquid region. It leads to higher temper‐ atures than in the case where only ice is considered. For instance it leads to a steady state

*Transient State*. The influence of phase transition on the transient state is now addressed. The temperature and heat flux solution of the heat conduction problem are plotted (see Figure 1) to give more insight on the physics that take place. It can be seen that the temperature reaches the melting point after 0.8s. Up to that point a single phase solution to the heat conduction

*tr*(0)= 2.8 °*C* = 276*K*

∂ *u* <sup>∂</sup> *<sup>x</sup>* (*x*<sup>0</sup>

W/cm2 calculated to impose T = 0°C in x = 100μm is used, the interface

+(*t*)). As a result, at steady state, the

(*L* ‐ *x*) + *Tc*. Tc is the temperature at x = L and symbolizes

∂ *u* <sup>∂</sup> *<sup>x</sup>* (*x*<sup>0</sup>

the two‐temperature fluidic interrogative platform.

interface.are then u(x0(t)) = 0 and*kwater*

temperature profile in the ice region is:

*kwater* (*x*0<sup>∞</sup> ‐ *x*) *if x* ∈ 0, *x*0<sup>∞</sup>

temperature in x = 0 of value *T* =*u∞*

(*L* ‐ *x*) + *Tc if x* ∈ *x*0∞, *L*

*tr*(*x*)=*u∞*(*x*)= *<sup>J</sup>*

*kice*

for *x* ∈ *x*0*∞*, *L* :*u∞*

the cooler's temperature.

If a heat flux J=1.2\*104

state solution is:

*u*∞ *tr*(*x*)= *<sup>J</sup>*

*u*∞ *tr*(*x*)= *<sup>J</sup> kice*

**2. Opto‐electro‐reconfigurable‐microchannel (OERM)**

C needs

**Figure 1.** Evolution of temperature and heat flux profiles in the first 5s after heating starts.

Unfortunately, *τmelting <sup>ʹ</sup>* is much longer than the characteristic transient times calculated previously. Because meanwhile the temperature profile will evolve towards homongeneiza‐ tion of the heat flux, it entails that the real melting time will probably be longer. At this point, for better understanding and assessment of the problem, more qualitative insight is required. When melting starts, it creates an interface at the melting temperature that gradually pro‐ gresses towards its steady state location. The speed of this progression is determined by the coupling of the latent heat of fusion and the derivative of the heat flux across the interface. Because the latent heat is much higherthan the heat capacity, phase transition is a much slower process than the increase oftemperature.It also imposes a fixed temperature. As a result, phase transition thwarts the heat flux homogenization process of heat conduction across the melting interface.The characteristic transienttime for an ice region of 1mm anda waterregion of 100μm do not depend on the heat flux and are respectively *τ ice* = 0.4*s* and *τ ice* = 0.03*s*. When comparing to *τmelting <sup>ʹ</sup>* , it seems a reasonable assumption to consider that steady state is reached in a negligible time with regard to melting, both in the liquid region and the ice region. If the interface is represented by a small interval, the heat flux difference across such interval can be assessed. As a consequence, it turns out that the overall melting time can be calculated rather precisely using the following method. The overall melting region can be discretized in small intervals that each correspond to a value d(t). For each d(t), the steady state profiles in the water and in the ice regions are estimated. The temperature in the liquid region will be described as *uwater*(*x*, *<sup>t</sup>*)= *<sup>J</sup> kwater* (*d*(*t*) ‐ *x*) while the temperature in the ice region will be *uice*(*x*, *t*)=*Tc L* ‐ *x <sup>L</sup>* ‐ *<sup>d</sup>* (*t*) ‐ *Tc*. Consequently the inward heat flux is J while the outward heat flux is *<sup>J</sup> out*(*d*(*t*))= *kiceTc <sup>L</sup>* ‐ *<sup>d</sup>* (*t*) . It entails that the melting time for each interval is *δτ ʹʹ* <sup>=</sup> *Hε <sup>J</sup>* ‐ *<sup>J</sup> out*(*<sup>d</sup>* (*t*)) , where *ε* is the size on an interval. Since the interface progression slows down, the interval can be discretized more finely near the steady state position. Here, *ε* = 10*μm* on the first 80μm, then *ε* = 1*μm* from x = 80μm to x = 99μm and finally *ε* = 100*nm*from x = 99μm to x = 100μm. The overall time calculated through this method gives an estimate of the melting duration of a 100μm channel with a precision of 100nm. It gives

Finally the case of freezing is also noteworthy. Microchannel freezing indeed happens at the very beginning of the cooling process: the temperature reaches 0°C within 0.1s. As a result, the temperature profile in the ice still imposes a high heat flux through Fourierʹs law while the heat flux at x = 0 is now zero. Since phase transition blocks the temperature evolution in the ice by fixing the temperature at the freezing point, freezing occurs at a heat flux value close to the highest one in the device: J. The freezing time can therefore be evaluated as *τ ice* = 0.4*s*. Steady state will be reached two seconds later according to the last section, hence after about 2.5s. The times required to melt a microchannel and freeze a microchannel have very different values; this is due to their timing within the transient state. Melting happens at the end, when the derivative of the heat flux is already close to zero: little heat is available for phase transition. On the contrary freezing occurs at the very beginning of the transient phase and thus benefits from high heat flux derivatives: freezing times are short. This observation bolsters the strategy for fast melting described hereinbefore that consists in using a high heat flux at the beginning, and to decrease its value when the desired microchannel size is obtained. Indeed, melting will start earlier in the transient phase providing higher heat flux gradients for phase transition. Those higher heat flux gradients concentrate across the interface when heat flux homogenizes in the liquid and solid zones. Besides, by reducing heat flux abruptly, the slow convergence of the interface to its steady state position is skipped. Therefore, since varying the heat flux offers schemes for fast melting, this analysis tends to prove that fast reconfiguration can be achieved. Additionally, it clearly proves that melted microchannels will be stable. Finally, the question of flows has not yet been addressed. A fluid entering a channel will thermally interact with its environment. It will probably have a higher temperature and could cause the channel shapes to vary. Nevertheless, since the transient time for small volumes is very short this issue

System Integration of a Novel Cell Interrogation Platform 303

can probably be easily addressed. It should not be a major concern.

proved the best candidate.

Theoretical models and simulations have proven the feasibility of microchannel reconfigura‐ tion by local melting. Preliminary experiments were also performed that corroborated the analysis and demonstrated liquid transport in such channels. They involved patterned metal electrodes that would locally thaw ice by resistive heating. Further work has characterized determined the output power of the phenomenon used in OERM, optoelectronic heating [10]. Building on those findings, we dealt with the complete platform for optoelectronic reconfig‐ urable microchannels, the OERM platform. OERM consist in locally melting microchannels in a frozen working media with light patterns. Since it is a reversible process, microchannels can reconfigure when the actuating patterns change. Figure 3 illustrates this principle. The previous paragraphs have reported the power needed for such task: direct melting by illumination would require high powerlight sources. There are great advantages to low‐power light for parallel manipulation, flexibility, and device integration. OERM thus rely on a transduction mechanism where low‐power light is converted into high‐power heat; it is achieved via optoelectronics and Joule effect. Therefore, such technology is built on the synergetic harnessing of several physical phenomena: light transmission, transduction of light into electrical currents, conversion of electrical currents into heat, melting, and micro flows: hence the complexity of the OERM platform. At its core lies a photoconductive material. Because of its high light absorption and very low thermal conductivity compared to other photoconductive materials such as crystalline silicon, hydrogenated amorphous silicon has

$$
\boldsymbol{\pi}\_{\text{melting}}^{\cdots} = \sum \boldsymbol{\delta} \,\boldsymbol{\tau}^{\cdots} = 181 \,\mathrm{s} = 3 \,\mathrm{min}.
$$

The movement of the solid liquid interface is plotted in Figure 2. When *τmelting ʹʹ* is added to the transient time, namely 2s, it becomes the time that a 100μm channel needs to reach stability with a precision of a hundred nanometers. The times to melt smaller regions with the same heat flux are also noteworthy, even thought the steady state melted region will still be a hundred micrometers long. For instance 50μm are melted in 18s, while 10μm are melted in 3s. Besides, 90μm and 99μm are reached respectively in 56s and 106s. Interestingly, the first approximation *τmelting <sup>ʹ</sup>* = 119s was not that far‐fetched in comparison. Moreover, it means that two third of the time are spent melting the last 10μm. Varying the heat flux could therefore shorten melting times tremendously. For instance a high heat flux can first be used to reach the desired size in a short time; it can then be decreased to the value corresponding to the steady state related to that size.

**Figure 2.** Evolution of the location of the water/ice interface with time.

Finally the case of freezing is also noteworthy. Microchannel freezing indeed happens at the very beginning of the cooling process: the temperature reaches 0°C within 0.1s. As a result, the temperature profile in the ice still imposes a high heat flux through Fourierʹs law while the heat flux at x = 0 is now zero. Since phase transition blocks the temperature evolution in the ice by fixing the temperature at the freezing point, freezing occurs at a heat flux value close to the highest one in the device: J. The freezing time can therefore be evaluated as *τ ice* = 0.4*s*. Steady state will be reached two seconds later according to the last section, hence after about 2.5s. The times required to melt a microchannel and freeze a microchannel have very different values; this is due to their timing within the transient state. Melting happens at the end, when the derivative of the heat flux is already close to zero: little heat is available for phase transition. On the contrary freezing occurs at the very beginning of the transient phase and thus benefits from high heat flux derivatives: freezing times are short. This observation bolsters the strategy for fast melting described hereinbefore that consists in using a high heat flux at the beginning, and to decrease its value when the desired microchannel size is obtained. Indeed, melting will start earlier in the transient phase providing higher heat flux gradients for phase transition. Those higher heat flux gradients concentrate across the interface when heat flux homogenizes in the liquid and solid zones. Besides, by reducing heat flux abruptly, the slow convergence of the interface to its steady state position is skipped. Therefore, since varying the heat flux offers schemes for fast melting, this analysis tends to prove that fast reconfiguration can be achieved. Additionally, it clearly proves that melted microchannels will be stable. Finally, the question of flows has not yet been addressed. A fluid entering a channel will thermally interact with its environment. It will probably have a higher temperature and could cause the channel shapes to vary. Nevertheless, since the transient time for small volumes is very short this issue can probably be easily addressed. It should not be a major concern.

precisely using the following method. The overall melting region can be discretized in small intervals that each correspond to a value d(t). For each d(t), the steady state profiles in the water and in the ice regions are estimated. The temperature in the liquid region will be

Micro-Nano Mechatronics — New Trends in Material, Measurement, Control, Manufacturing and Their Applications in

*<sup>L</sup>* ‐ *<sup>d</sup>* (*t*) . It entails that the melting time for each interval is *δτ ʹʹ* <sup>=</sup> *Hε*

The movement of the solid liquid interface is plotted in Figure 2. When *τmelting*

is the size on an interval. Since the interface progression slows down, the interval can be discretized more finely near the steady state position. Here, *ε* = 10*μm* on the first 80μm, then *ε* = 1*μm* from x = 80μm to x = 99μm and finally *ε* = 100*nm*from x = 99μm to x = 100μm. The overall time calculated through this method gives an estimate of the melting duration of a

transient time, namely 2s, it becomes the time that a 100μm channel needs to reach stability with a precision of a hundred nanometers. The times to melt smaller regions with the same heat flux are also noteworthy, even thought the steady state melted region will still be a hundred micrometers long. For instance 50μm are melted in 18s, while 10μm are melted in 3s. Besides, 90μm and 99μm are reached respectively in 56s and 106s. Interestingly, the first

two third of the time are spent melting the last 10μm. Varying the heat flux could therefore shorten melting times tremendously. For instance a high heat flux can first be used to reach the desired size in a short time; it can then be decreased to the value corresponding to the

*<sup>ʹ</sup>* = 119s was not that far‐fetched in comparison. Moreover, it means that

*<sup>L</sup>* ‐ *<sup>d</sup>* (*t*) ‐ *Tc*. Consequently the inward heat flux is J while the outward heat flux is

(*d*(*t*) ‐ *x*) while the temperature in the ice region will be

*<sup>J</sup>* ‐ *<sup>J</sup> out*(*<sup>d</sup>* (*t*)) , where *ε*

*ʹʹ* is added to the

described as *uwater*(*x*, *<sup>t</sup>*)= *<sup>J</sup>*

*L* ‐ *x*

*uice*(*x*, *t*)=*Tc*

*τmelting* ʹʹ =∑ *δτ* ʹʹ

*<sup>J</sup> out*(*d*(*t*))= *kiceTc*

Biomedical Engineering

302

approximation *τmelting*

steady state related to that size.

*kwater*

100μm channel with a precision of 100nm. It gives

**Figure 2.** Evolution of the location of the water/ice interface with time.

= 181*s* ≈3*min*

Theoretical models and simulations have proven the feasibility of microchannel reconfigura‐ tion by local melting. Preliminary experiments were also performed that corroborated the analysis and demonstrated liquid transport in such channels. They involved patterned metal electrodes that would locally thaw ice by resistive heating. Further work has characterized determined the output power of the phenomenon used in OERM, optoelectronic heating [10]. Building on those findings, we dealt with the complete platform for optoelectronic reconfig‐ urable microchannels, the OERM platform. OERM consist in locally melting microchannels in a frozen working media with light patterns. Since it is a reversible process, microchannels can reconfigure when the actuating patterns change. Figure 3 illustrates this principle. The previous paragraphs have reported the power needed for such task: direct melting by illumination would require high powerlight sources. There are great advantages to low‐power light for parallel manipulation, flexibility, and device integration. OERM thus rely on a transduction mechanism where low‐power light is converted into high‐power heat; it is achieved via optoelectronics and Joule effect. Therefore, such technology is built on the synergetic harnessing of several physical phenomena: light transmission, transduction of light into electrical currents, conversion of electrical currents into heat, melting, and micro flows: hence the complexity of the OERM platform. At its core lies a photoconductive material. Because of its high light absorption and very low thermal conductivity compared to other photoconductive materials such as crystalline silicon, hydrogenated amorphous silicon has proved the best candidate.

but also it can evolve on its own thanks to a feedback control loop. This is a real breakthrough

System Integration of a Novel Cell Interrogation Platform 305

Last year a first version of the optoelectronic chip was manufactured that led to the first demonstrations of local melting controlled by low‐power light images. The set‐up consisting of the electrical controls, the cooling system and the optical projection system was assembled

for microfluidics and offers a whole new range of possible applications.

and optimized. It paved the way to this year's accomplishments.

**Figure 4.** Working principle of OERM

**Figure 5.** OERM integrated with plamsonic sensor platform

**Figure 3.** OERM principle: microchannel reconfiguration controlled by light patterns. 1) Working media is frozen. 2) A first light pattern is displayed, which creates a corresponding microchannel; through external pumping green particles (channel end, left-hand side of image) and purple particles (channel first branch, right-hand side of image) flow through. 3) A second light pattern is displayed; the microchannel reconfigures accordingly, and the purple particles (end of new channel) are directed towards another outlet.

OERM is a new microfluidic platform that can provide incomparable flexibility in the planar manipulation of bio‐molecules in microchannels. Thanks to this new technology any micro‐ fluidic network can be manufactured and reconfigured within seconds. It will be used on the plasmonic sensor platform for cell, drug and sensing probe delivery (cf. Figure 5).

A schematic overview of the optoelectronic reconfigurable microchannel platform is shown in Figure 4. An optoelectronic chip is fabricated by successively depositing on a glass substrate a transparent conductive layer made of indium tin oxide (ITO), a photoconductive layer made of hydrogenated amorphous silicon (a‐Si:H), and a highly reflective conductive layer made of titanium, platinum and silver or gold. A liquid layer is laid upon the optoelectronic chip and is confined between the chip and a cover slip. The system is placed on a transparent cooler to freeze the liquid. A voltage bias is then applied between the two conductive layers of the optoelectronic chip resulting in a constant electric field across the photoconductive layer. A light pattern is projected on the optoelectronic chip through the transparent cooler. It is absorbed by the photoconductive layer (a‐Si:H). In the dark parts of the light pattern a‐Si:H behaves as an insulator, hence no electrical current flows across it. On the contrary in the bright parts of the light pattern local currents flow across a‐Si:H. Those currents thus create local Joule heating. This process is called optoelectronic heating and is schematically represented (cf. Figure 4). Setting the local Joule heating to the right power triggers local melting in the frozen liquid, thus forming microchannels corresponding to the light pattern. When light stops shining on a microchannel, local Joule heating stops and the fluid in the microchannel freezes thus closing the microchannel. If the light pattern has reconfigured, other microchannels will open in otherlocations on the chip. Flows can be produced in the microchannel network using an external pumping system so other phases or particles can be manipulated in the network. The melted liquid is kept as the main phase to allow the melted microchannels to freeze back when light patterns disappear or reconfigure.

The key novel feature of this technology is that the light pattern can be changed at will, hence making the microfluidic network reconfigure instantaneously. This potentially allows for any dynamic two‐dimensional design of the microchannel network. The projected pattern is controlled by a computer, which entails that not only can the light pattern be pre‐programmed, but also it can evolve on its own thanks to a feedback control loop. This is a real breakthrough for microfluidics and offers a whole new range of possible applications.

Last year a first version of the optoelectronic chip was manufactured that led to the first demonstrations of local melting controlled by low‐power light images. The set‐up consisting of the electrical controls, the cooling system and the optical projection system was assembled and optimized. It paved the way to this year's accomplishments.

**Figure 4.** Working principle of OERM

**Figure 3.** OERM principle: microchannel reconfiguration controlled by light patterns. 1) Working media is frozen. 2) A first light pattern is displayed, which creates a corresponding microchannel; through external pumping green particles (channel end, left-hand side of image) and purple particles (channel first branch, right-hand side of image) flow through. 3) A second light pattern is displayed; the microchannel reconfigures accordingly, and the purple particles

Micro-Nano Mechatronics — New Trends in Material, Measurement, Control, Manufacturing and Their Applications in

OERM is a new microfluidic platform that can provide incomparable flexibility in the planar manipulation of bio‐molecules in microchannels. Thanks to this new technology any micro‐ fluidic network can be manufactured and reconfigured within seconds. It will be used on the

A schematic overview of the optoelectronic reconfigurable microchannel platform is shown in Figure 4. An optoelectronic chip is fabricated by successively depositing on a glass substrate a transparent conductive layer made of indium tin oxide (ITO), a photoconductive layer made of hydrogenated amorphous silicon (a‐Si:H), and a highly reflective conductive layer made of titanium, platinum and silver or gold. A liquid layer is laid upon the optoelectronic chip and is confined between the chip and a cover slip. The system is placed on a transparent cooler to freeze the liquid. A voltage bias is then applied between the two conductive layers of the optoelectronic chip resulting in a constant electric field across the photoconductive layer. A light pattern is projected on the optoelectronic chip through the transparent cooler. It is absorbed by the photoconductive layer (a‐Si:H). In the dark parts of the light pattern a‐Si:H behaves as an insulator, hence no electrical current flows across it. On the contrary in the bright parts of the light pattern local currents flow across a‐Si:H. Those currents thus create local Joule heating. This process is called optoelectronic heating and is schematically represented (cf. Figure 4). Setting the local Joule heating to the right power triggers local melting in the frozen liquid, thus forming microchannels corresponding to the light pattern. When light stops shining on a microchannel, local Joule heating stops and the fluid in the microchannel freezes thus closing the microchannel. If the light pattern has reconfigured, other microchannels will open in otherlocations on the chip. Flows can be produced in the microchannel network using an external pumping system so other phases or particles can be manipulated in the network. The melted liquid is kept as the main phase to allow the melted microchannels to freeze back

The key novel feature of this technology is that the light pattern can be changed at will, hence making the microfluidic network reconfigure instantaneously. This potentially allows for any dynamic two‐dimensional design of the microchannel network. The projected pattern is controlled by a computer, which entails that not only can the light pattern be pre‐programmed,

plasmonic sensor platform for cell, drug and sensing probe delivery (cf. Figure 5).

(end of new channel) are directed towards another outlet.

Biomedical Engineering

304

when light patterns disappear or reconfigure.

**Figure 5.** OERM integrated with plamsonic sensor platform
