ProceedingsPhysics with

Industry 2016Lorentz Center Leiden, the Netherlands, 21- 25 November

Proceedings Physics with Industry 2016

Lorentz Center Leiden, the Netherlands, 21-25 November 2016

1

Colophon

Text Participants and organisation workshop.

Cover photo Foto credit: Supernova Studio’s

Design and production NWO, Utrecht, The Netherlands.

Thanks to The organisation is particularly grateful for the excellent service and facilities of the Lorentz Center, the effort of the senior researchers (from the preparation phase onwards) and the enthusiastic contribution of all participants during the week. The workshop 'Physics with Industry' was organised by NOW in collaboration with the Lorentz Center. The event was funded by the Lorentz Center (which is partly funded by NWO) and the participating companies.

2

Proceedings Physics with Industry 2016

Table of contents

Event report 4

Case 1 : Dunea 6

Case 2 : Shell 21

Case 3 : Sensortag 27

Case 4 : Lievers 33

Case 5 : Philips 38

3

Event report Physics with Industry 2016

From charged mesh nebulisers to purifying water Michiel van den Hout, Jeroen van Houwelingen, Melvin Kasanrokijat, Maria Sovago, Xavier Weenink

Description and aims

In this year's edition of the Physics with Industry workshop, 35 researchers worked on five real-world

industrial problems during five consecutive days. Dunea, Lievers Holland, Philips, SensorTags Solutions

and Shell participated with an industrial case.

The industrial cases were selected by a scientific committee after an open call during which companies

could submit a case. All of the industrial case owners declared that the proposed solution of the

workshop week really helped them to develop their case further. The submitted cases ranged from

testing theoretical or hypothetical opportunities to improving existing technologies. Surprisingly, the

most practical case, dealing with the compacting of concrete, showed the most out-of-the-box solution,

but some real experiments had to be done to get there. In another case, where the issues in the

generation of ultrapure water droplets using a mesh nebuliser were investigated, the solution emerged

from theoretical analysis. A third case used modelling as a tool to obtain a solution.

Format

All of the cases were coached by an academic and an industrial supervisor. This guaranteed the

scientific quality and the applicability of the solution. The participants enjoyed the workshop due to the

scientific challenge but also through experiencing how industrial problems are solved and how

companies work. A few comments from the participants:

'I had the opportunity to experience for a week what it feels like to work for a company. It made me

realise that my knowledge can actually be applied to real problems, which is very reassuring.'

'I have seen what the benefits and dynamics of teamwork are and I learned a lot about

practical/industrial problems.'

'I learned a lot about industry life. As a PhD student you are used to academic life and industry can be

scary. But this workshop and our case showed that industry can go hand in hand with science.'

The workshop was held from 21 to 25 November 2016 at the Lorentz Center in Leiden. The project is a

joint collaboration between FOM, Technology Foundation STW and the Lorentz Center. The findings and

suggested solutions will be published online by the end of January 2017.

4

Presentation and award ceremony on Friday (Credit: Auke Planjer, Lorentz Center)

Participants Physics with Industry 2016 (Credit: Eline Pollaert, Lorentz Center)

5

1

Responsible dosing of iron(II)sulphate in

a drinking water source; in control of

coagulation K. Akius1*, D. Davidovikj1, E. Hekhuizen2 , T. Knol2, J.T. Padding3, V. T. Tenner3, R. Thorat,

E.C.I. van der Wurff4

Abstract In these proceedings we describe the work performed for the Dunea case during Physics

with Industry 2016 at the Lorentz center in Leiden. The Dunea case centered around the

dosing of iron(II)sulphate in a side branch of the river Maas, which is used as a source

for drinking water. This dosing is done to remove phosphate from the river water, which

prevents blooming of harmful blue algae. After an analysis of the problems of the current

set-up, we give suggestions for better diagnostication and possible improvements.

Introduction 3

Current implementation and problems 4

Considerations 6

Approaches 9

Geometric optimization of the pipes and nozzles 9

Mechanical valves 11

Electronic valves 13

Active clogging prevention 14

Conclusions 14

1 Leiden University 2 Dunea 3 Delft University of Technology 4 Utrecht University * All authors contributed equally to this work

6

2

Introduction Dunea supplies drinking water to 1.2 million people in the South Holland region in the

Netherlands. One of the sources of the water production is from Managed Aquifer Recharge

which is a process in which water is artificially resupplied or recharged in natural wells. Dunea

uses a side branch of the river Meuse called the Afgedamde Maas as a source for recharging

natural wells in the form of natural sand dunes in the Katwijk area. The water taken from the

Afgedamde Maas needs to be treated before resupplying to the dunes. One reason for this is

high content of nutrients from the agricultural area upstream from the sourcing site causing toxic

blue algae bloom in summers. The blue algae grow particularly well in the presence of high

concentrations of phosphate. This is dangerous because when blue algae decompose, they

excrete cyanotoxins that are harmful to both humans and animals. Thanks to environmental

regulation, both in domestic phosphorus pollution (from detergents etc.) as well as in agriculture,

the phosphate levels in the Meuse has been declining steadily over the past few decades. Data

supporting this trend is presented in Figure 1.

Figure 1: Left: blue algae bloom on Lake Crystal, Minnesota (USA). Right: decreasing concentration of

phosphate in the Maas river from 1976 to 2014 in mg/L.

Still, it is necessary to reduce to amount of phosphate in the river water to prevent algae

blooms. One common strategy is to use iron or aluminium to bind phosphates in the water5. In

this direction Dunea has implemented a iron(II)-sulphate dosing system that has kept their

reservoir free from blue algae blooms since 1997. The current implementation is due for an

overhaul as the requirements have changed and the current pipe installation is reaching the end

of its lifetime. Naturally, adding large amounts of chemicals can have adverse effects on the

water quality and wildlife in the water so environmental concerns are important in the design.

Running the plant at low efficiency is also costly in terms of chemicals and power consumption,

5 Parmar et al., ARPN Journal of Engineering and Applied Sciences, 2011

7

3

further adding to the environmental footprint of the plant. Optimizing these factors is the main

motivation for this project.

Figure 2: Schematic displays of the pipe that is currently used. a) Cross-sectional view: a small pipe for

the dosing of iron sulphate with a diameter of 28 mm is embedded in a bigger pipe for the air with a

diameter of 110 mm. Both pipes have nozzles protruding from the top. b) Top view of one group of

nozzles. Every group of nozzles contains three air nozzles with a diameter of 2,4 mm and one smaller

iron sulphate nozzle with a diameter of 1.5 mm. c) The total pipe is about 100 meters long and contains

70 of such groups of nozzles.

Current implementation and problems

The current installation consists of a pipe, approximately 100 m long along the river bottom, a

cross section can be seen in Figure 2 a). The pipe carries air as well as liquid phase iron(II)-

sulphate dose in a smaller pipe inside. A schematic image of the details of the the nozzles and

the distribution of the nozzles are shown in Figure 2 b) and c) respectively. The purpose of the

air is twofold, i) aerating the river to compensate for the oxygen consumption from the reaction

4 𝐹𝑒2+ + 𝐻3𝑂+ + 𝑂2 ⇒ 4 𝐹𝑒3+ + 6 𝐻2𝑂

and ii) enhancing the mixing of the constituents.

One of the issues the current system has is that the air nozzles tend to clog a few times

every year, which has associated downtime and maintenance. The operators notice the

clogging of the nozzles by an increase in the pressure of the air and iron sulphate pipes. The

pressure profiles in both pipes are displayed in Figure 3. For the air pipe the data is

unambiguous: the pressure slowly increases over time until the situation becomes

unmanageable and a diver is hired to clean the air nozzles. Cleaning of the nozzles leads to a

drop of the pressure; one such cleaning in June of 2016 can be seen in Figure 3.

However, the pressure profile in the iron sulphate pipe seems to show a global

decreasing behavior from March until July. This could be a consequence of the fact that the

viscosity of most solutions decreases as their temperature increases, something which happens

exactly in these months. We estimated the associated pressure drop and it seems to be in the

right value range based on the data depicted in Figure 3. However, this is an approximation and

at this point it is difficult to draw any conclusions about the iron-dosing pipes and if, when and

8

4

why they clog. With the data at hand it seems it is mostly the air pipes that clog so the focus of

our efforts will be on the clogging of the air nozzles.

Figure 3: Left: the pressure (kPa) in the air (orange) and iron sulphate (blue) pipes from January 2016 to

November 2016. The sudden drops to zero correspond to the switching off of the pumps. Right: zoom in

on the data of the air pump. The pressure increases to a maximum value around the beginning of June,

when a diver unclogged the nozzles. After that the pressure starts to increase again.

An additional issue is that the system uses a lot of energy, particularly the air pumps

consume 21 kW. All iron sulfate dissolves in water to give the same aquo complex [Fe(H2O)6]2+.

When mixed with air, iron (II) sulphate solution further gives iron(III) sulphate and iron oxide

according to the reaction:

12 FeSO4 + 3 O2 → 4 Fe2(SO4)3 + 2 Fe2O3

Iron (III) sulphate further gives iron (III) charged molecules to form Iron hydroxides

Fe(OH)3 with water: 3Fe2(SO4)3 → 2Fe3+ + 9(SO4)-

Ferric hydrate (FeSO4.7H2O) contains 55 g per mole of iron. If it is assumed that all iron from the iron sulphate solution oxidizes, 64 g of oxygen molecules is required to convert 56 g of iron to form iron phosphate FePO4. As oxygen is one fifth of air, 320 g of air is required for 56 g of iron to Fe3+ and further to iron phosphate. If we consider to dose, 25 tonnes of iron sulphate solution of which 5 tonnes is iron, it would require 30 tonnes of air. The density of air (at atmospheric conditions) is 1.205 kg/m3. The maximum flow rate at which pump can inject air is 0.2 m3/s, so the pump can inject air at 0.245 kg/s. Therefore to inject 30 ton of air at a maximum power of the pump, it will take ~35 hours, assuming all air is mixed with the iron.

To summarize the objectives of the project we want to lower the

● Chemical dosage;

9

5

● Downtime;

● Maintenance costs;

● Power consumption.

Considerations

The objective of the water treatment is to keep phosphate levels low enough to keep the water

clear and divest the water of nutrients to avoid blue algae blooms that could be detrimental for

water production. Many considerations have to be taken into account in this problem as the

complexity is very high.

The Afgedamde Maas is a dead river arm of the Maas. The intake is 6km from the

entrance, the dosing facility is 4 kilometers from the intake. Hence there is much space for the

reactions for the water treatment. The net flow is 10 meters per day and is caused by the pump

at the end of the river. The main waterflow is caused by the tides. This tidal wave moves the

water 2.5 kilometers in 6 hours. This flow provides significant passive mixing power. There are

two distinct passive mixing processes: vertical and lateral mixing. For both processes we can

associate a specific length scale depending on the dimensions of the river, and a time scale

depending on the flow speed. Passive vertical mixing in the river occurs on length scales of

around 70 meter, which takes up to 6 hours. This time is approximated to be longer than the

active period of the iron-salt mixture, hence active mixing is necessary to ensure sufficient

mixing. Careful chemical analysis of the dynamics of the reactions involved is necessary to

make any confident conclusions.

10

6

Figure 4: Liquid film coefficient KL in cm/hr as a function of the diameter dB of the air bubbles coming from

a air nozzle. Optimal aeration of the water, corresponding to the highest value of KL, occurs at a bubble

diameter of about 2,5 mm. Empirically, this corresponds to typical nozzle diameters of 0.1 to 0.3 mm.

As mentioned earlier the injected air is responsible for mixing and for aeration, or

oxygenation of the river. The size of the air bubbles is critical for the diffusion of the air into the

water. This is a way to control how efficient the aeration is in terms of the amount of gas that is

actually absorbed in the water, as can be seen in Figure 4. A large part of the expected savings

are associated with costly operation of the aeration pumps. Stoichiometrically optimizing the

aeration to obtain a minimum oxygen level for which the desired reactions will still occur

sufficiently is probably crucial to minimize air pumping. This will in turn give a boundary

condition for the minimum pumping, with a given bubble size. Finally this amount of pumping

still has to be enough to mix the solutions sufficiently.

The iron is put in the river in the following way. Dunea dissolves commercially available

FeSO4 in water in tanks on the river bank in batches of 25 ton of salt. This yields Fe2+ and SO42-

ions. When this fluid is injected into the water stream, the Fe2+ reacts with O2 to form Fe3+. The

iron oxidation is a rather slow process with a typical time between 1 and 200 minutes. Only the

Fe3+ ions can react with PO43- to form solid FePO4. Besides the phosphate capture, Fe3+ also

reacts with OH-. This process is very fast, and hence a more than stoichiometric ratio should be

added in order to have phosphate removal. Such iron-hydroxides have the beneficial properties

that they form flocks, who coagulates and removes sediments from the stream.

Figure 5: Three mixing conditions of Fe3+ in the river.

The phosphate removal process strongly depends on both the dosing and the mixing

properties. There are three regimes, as indicated in Figure 5. When the Fe3+ concentration is

too low, no phosphate is removed, only sediments are removed. On the other hand, when the

iron solution is not mixed, the Fe3+ concentration is very high and hence the phosphate removal

is good, but only in a small volume. The optimal mixing strategy is thus dependent on the dose

and dosing speed. More numerical modelling is advised in order to optimize the total injected

dose.

Additionally, there are many considerations that are specific for the conditions of the

dosing station, everything from biomass content, flocculant size and biological activity to take

11

7

into account. In general, in winter algae are not growing and hence the dosing can be much

lower, hence the dosing of iron is only 20% of the summer time dosing, as shown in Figure 6.

During the oxidation of Fe2+ to Fe3+ oxygen is removed from the water. This oxygen loss

has to be compensated, which is currently implemented by bubbling compressed air through the

water. This supply of air also helps with mixing. Currently, the air supply is kept constant even

though the iron dosing is only a fraction in winter compared to summer. Adjusting the air supply

to the dosing presents a large potential saving.

Due to the aforementioned complexity of these systems and the data on the actual

pumping system available, most of these questions cannot be answered quantitatively from

theoretical approaches in physics alone. One can attempt engineering approximations, which is

what has been done in this workshop. Another promising approach would be to attempt looking

at these problems with chemistry method and analysis.

Figure 6: Planned dosing of Iron sulphate (kg) in the different weeks of 2016. The amount ranges from

less than 5000 kg in winter to 25000 kg in the summer. In the summer there is more biological activity due

to a higher average temperature of the water, resulting in a more pressing need to remove the phosphate

from the river water to prevent (blue) algae blooms.

Approaches

Disentangling all the variables and processes is a daunting task. An alternative approach to this

would be to simply measure the phosphate levels for different mixing and dosing methods and

pressures and ensure that the levels are acceptable. With this information in hand, more

substantial claims can be made. Nevertheless some general advice can be given.

We will first optimize the geometric design of the set-up, such that it is expected to clog

less frequently, or ideally not even at all. Depending on what the clogging mechanism is, a

backflow-preventing valve may need to be installed. These may be simple mechanical valves or

electronically operated valves. Despite all these measures, the nozzles may still clog after a

longer period of time. Therefore, we have also considered active cleaning options which

12

8

automatically unclog the nozzles. In the following all approaches will be described in some

detail.

Geometric optimization of the pipes and nozzles

From an engineering point of view, a geometric optimization of the pipe and nozzle dimensions

seems to be the most natural candidate. Can we think of simple changes we can make to the

pipes and nozzles, to prevent them from clogging as quickly as they do now?

The piping system is close to the end of its lifetime and needs to be replaced in the next

few years. Naturally, the system upgrade is an opportunity to also optimize the geometry of the

piping and the distribution of the nozzles. The hydrodynamic system is governed by a rather

simple equation:

ΔP= α LQ2T/(d5N2),

where ΔP is the pressure difference across the nozzles, L is the length of each nozzle, d is its

diameter, QT is the total flow through all the nozzles, N is the number of nozzles and α is a

constant related to the properties of the flow.

Based on this relation, there are a few parameters that can be tuned to get optimal

conditions: the length L, the diameter d and the number of nozzles N. We also need to take into

account a few constraints and, naturally, the goals of the optimization:

● To reduce dosing, we would like to reduce the minimum flow that the system can

handle (Qmin). Currently this value is: Qmin = 200 l/h and our goal is to reduce this at least

by a factor of 2.

● The maximum flow the system should be able to produce is fixed and has to do with

emergencies, i.e. outbreaks of blue algae. The maximum flow rate should be Qmax =

1000 l/h.

● To prevent clogging and/or reduce clogging maintenance costs, a combination of

reducing the number of nozzles and increasing the nozzle diameter would be a

possible direction.

● The minimum operating pressure of the pump should be kept well above 0.6 bar to

counteract the pressure of the water in the river at the depth of the pipe and prevent

water from entering the pipes.

To find the optimal solution satisfying the constraints, we ran calculations varying each of the

proposed parameters. In Figure 7 we summarize the results for both the FeSO4 and the air

nozzles.

13

9

Figure 7: A plot of the required pressure ranges for the pumps as a function of nozzle length and diameter

with the goal of extending the dynamic range of the flow. Left: dynamic range for the FeSO4 pump for a

flow range of 70-600 l/h. Right: dynamic range for the air pump for a flow range of 5000 - 9000 l/m. The

white space signifies that a solution could not be found for the given set of parameters and constraints.

As mentioned above, it is of interest to increase the diameter of the nozzle, assuming

that a larger diameter would lead to less clogging. Increasing the length of each nozzle, should

pose no issue in general. However, installing very long nozzles may yield them more brittle and

they may break when subjected to stronger turbulence. It is thus important that the nozzles are

kept reasonably short (< 40 cm) and as wide as possible, provided that they can supply the

minimal required flow with the minimum pressure difference that the pump can handle (in this

case, it is assumed that the pump can have a pressure resolution of at least 0.01 bar). From

Figure 7 (left panel) we can conclude that an FeSO4 nozzle with a length of 30 cm and a

diameter of 1.9 mm would be the optimal given the constraints. Using pressures between 0.6

and 2 bar, one should be able to cover the entire flow rate range for the standard operating

conditions (70 - 600 l/h). Doing this, the minimum possible flow would be reduced by a factor of

2, which would prevent overdosing in periods of low tide and would eliminate the need of

diluting the iron sulfate solution in the winter season.

A similar calculation for the air nozzles is shown in Figure 7 (right). In this case,

however, there is an additional consideration that we need to take into account. The efficiency

of the aeration depends strongly on the air bubble size, which in turn depends on the nozzle

diameter. The ideal bubble size for aeration according to literature is around 2.5 mm, which

would require nozzles of around 1mm in diameter. This is, of course, not optimal for the system

taking into account all the aforementioned considerations. There is a trade-off between

increasing the diameter to get less clogging and decreasing it to get smaller bubbles. Closest to

the optimal value would be a diameter of 4.2 mm which would require a very short nozzle (5

cm). It is important to note that such a configuration should be able to handle much smaller flow

rates (5000 l/h, compared to the standard 8000-9000 l/h) which, in periods of low dosage, could

significantly cut down on power consumption for pumping.

14

10

Mechanical valves

In case of an emergency, such as a pump failure or a power cut, the pressure in the iron dosing

pipe and air pipe may decrease to levels below 0.6 bar. This will allow river water to enter the

nozzles and pipes. This is undesired because river water in the air pipe will be difficult to

remove, and river water in the iron dosing nozzles may lead to initiation of the precipitation

mechanisms, and consequent clogging of the nozzles.

An obvious way to prevent backflow of river water into the system is to use backflow-

preventing valves. In industrial settings such valves, called check valves, are often installed in

pipe systems. However, we have a more specific need here: we need a valve which

automatically closes the exit of the nozzles as soon as the flow (either iron solution or air) is

stopped or in danger of reversing. This could theoretically be accomplished by a ball and spring

system as depicted in Figure 8 below. Under normal circumstances, the flow should be strong

enough to overcome the spring force.

Figure 8. Examples of mechanical backflow-preventing valves which close the exit of the

nozzles as soon as the flow through the nozzle stops or is in danger of reversing. Under normal

operation, the flow should be strong enough to overcome the spring force.

A back-of-the-envelope calculation can be made for the required spring constants.

Assuming that all momentum present in the fluid flow through a nozzle is just stopped by the

spring at a spring extension of the order of a few mm (the nozzle diameter), we find that a

typical flow force 𝐹𝑓𝑙𝑜𝑤 =1

2𝜌𝑣2𝐴 , with A the nozzle cross-sectional area, should be balanced

by a spring force 𝐹𝑠𝑝𝑟𝑖𝑛𝑔 = 𝑘𝑥, with k the spring constant and x the required spring extension,

which we take equal to the nozzle radius. For the air nozzles the flow is relatively strong (with air

velocities of the order of 102 m/s), which leads to a spring constant of the order of 102 N/m.

15

11

However, the flow velocity in the iron dosing nozzles is so low (of the order of 1 m/s) that the

spring constant for these nozzles would need to be of the order of 10-2 N/m. This is probably so

weak that any (temporary) turbulent flow in the river can also open such a valve. This is a

serious problem, because clogging is likely occur when river water can enter the iron dosing

nozzles.

In summary, the air nozzles may be automatically closed by installing mechanical

valves, but the iron dosing nozzles need a valve which does not depend on the momentum (or

pressure) of the iron solution at the nozzle exit.

Figure 9: Proposed wiring schemes for electrically controlled nozzle valves. Top: Separate

groups of nozzles for the center of the river bed and the sides. Bottom: alternate (odd/even)

configuration useful for intermittent dosing. Each color represents a group of nozzles.

Electronic valves

One approach to overcome this issue would be to install an electronic valve for each

nozzle. This has the obvious advantage that everything could be electrically controlled, so

during maintenance, power outages and even extremely low dosing, the valves could be closed

remotely, preventing backflow and cutting down the dosing completely. Such valves could also

reduce the chance of clogging by opening and closing more frequently. Another step forward

could be grouping the nozzles (i.e. valves) in two or more separate groups that can be

individually controlled. In such a way one could envision control schemes where the nozzles on

the sides of the river bed are more frequently opened than the ones in the center to compensate

for the lower dosing on the river sides (see Figure 9, top panel) or even a scheme where even

and odd nozzles are grouped separately, to provide for the option to halve the flow in period of

low dosing (Figure 9, bottom panel). The installation of electronic valves would also enable

implementing the intermittent dosing scheme more easily, as electrical pulses could be applied

to the nozzles such they are open only a certain percentage of the time. The major

disadvantage would be the installation cost of such a system and the fact that there may be

sensitive electronics under water.

16

12

Active clogging prevention

Several different approaches were suggested, ranging from temperature induced expansion of

the nozzles and piezo-induced vibration, to ultrasound dissolving of the clogs. However, all of

these seem to be unfeasible in the long term. One has to take account of the fact that the pipe

should be situated in the river for a rather long time. High tech solutions like piezo-induced

vibrations, or even local heating of the nozzles therefore seem to be far-fetched for the current

stage of the system.

Conclusions

Below we list some substantial advice on what questions should be answered and some nozzle

design and dimensions. The first two advice are meant to obtain a better characterization of the

system. The rest are are on options to optimize the system:

1) One important question to answer is what the nature of the substance is that clogs the

air nozzles. On-site human observations describe it ranging from a “fluffy” to a “hard”

substance. X-ray diffraction at several sites with similar clogging problems have resulted

in the identification of calcium sulfate and calcium phosphate as being the clogging

agent.6 Other data indicates it is mainly ferrihydrite.7 Due to this uncertainty, it is

desirable to harvest some of the clogging material and investigate what it is made of.

2) More systematic measurements have to be done to characterize the system better. At

the very least the phosphate, oxygen and iron concentration before and after the dosing

station should be measured at different dosing conditions. It is very possible that the air

dosing can be halved in winter time but no claims can be made without knowing the

phosphor level responds to this. In depth knowledge about the chemistry and mixing is

required.

3) At the moment the air pump has only one power option. Using an air pump that has

different power levels will allow a reduction of the power level in the winter period. In this

period less iron sulfate is dosed and consequently less oxygen is removed from the

water. Therefore, a pump with a lower power suffices. This will reduce the energy costs

in the winter period considerably because the necessary power scales with the

volumetric rate to the power of 3.

4) If significant improvements in the dosing are to be obtained, it would be advisable to

investigate the possibility to use a control system (feedback or feed forward)

investigated. This is common practice to implement in wastewater plants and should in

6 J.A. Müller, W.C. Boyle, and H.J. Pöpel, “Water Quality Management Library (Volume 11). Aeration: principles and practices”, CRC Press (2002). 7Medina et al., “Iron-hydroxide clogging of public supply wells receiving artificial recharge”, Hydrogeology

Journal (2013).

17

13

principle be possible to implement here too8,9. One difficulty with the control system route

is to find a feasible input parameter to feedback on. Just using phosphate levels is

insufficient as the phosphor level goes down when there is an algae bloom in the river.

This means that during a algae bloom a feedback system would lower the dosage

instead of increasing it. However, if one can model the algae risk well or find a suitable

marker or combination of markers that one can measure, it should be feasible. In 2016

Dunea commissioned a report that produced a model for the phosphate levels and we

see that it is quite accurate but it is indeed less accurate in summer. The possible

savings in phosphate levels can be approximated by comparing the dosing levels in

Figure 6 with the actual measured values of phosphor in Figure 10. As there is a large

discrepancy between the two, there should be considerable savings possible.

Alternatively, measurement of the phosphor level could be done in the river Maas and

use this as the feedback reference. Because of its higher flow velocities it is much less

likely to develop algae blooms and it therefore will not have the same issue with

misleading phosphor levels.

Figure 10: Figure from a commissioned report from the Water Research Institute in 2016. Using a simple

model they make predictions on the phosphate level in blue to be compared with the measured values in

red. The dashed blue line is the actual dosing and the dashed green line is the measured phosphor level

in the Maas. If data like this is available on at least monthly basis, feedback should be feasible.

8 http://www.worldpumps.com/view/316/control-of-chemical-dosing-in-wastewater-treatment/ Retrieved on

Nov 24, 2016 9

https://www.epa.ie/pubs/advice/drinkingwater/Drinking%20Water%20Advice%20Note%2015%20WEB.pdf

18

14

5) At the moment the nozzles are made of arnite (a high strength and rigid plastic).

Clogging of nozzles may be enhanced by a sufficiently high strength of adhesion

between sediments / precipitates and the nozzle material. It is advised to investigate the

performance of different nozzle materials (or coatings) with relatively low strength of

adhesion, such as PTFE (teflon) or ceramics. To optimize this, the clogging agent needs

to be analyzed as per point 1.

Figure 11: Different pipe designs. a) Original pipe design. b) Pipe design in which the air nozzle (grey) is

longer than the iron sulphate nozzle (yellow). c) Similar to b). d) Rotated design in which the air nozzle

pipe is situated below the iron sulphate nozzle, hence propelling the iron upwards and preventing

clogging of the air nozzle.

6) Replacing the current piping with new dimensions. Figure 12 shows the missing nozzles

in the current system. A missing nozzle gives poor performance down the pipe from a

clogged nozzle, as would be expected from a large pressure drop as a result from a

missing nozzle. A new design for a nozzle can be seen in Figure 11. The angle on the

air pipe is meant to eliminate backflow from water into the air pipe when the pump is

turned off. Further, if possible a check valve should be implemented, either on the main

pipe to keep backflow from occurring or (if feasible with the choice of pressure range) on

the nozzle itself.

19

15

Figure 12: Measurements on the clogging of the different groups of nozzles, numbered from one side of

the bank (1) to the other bank (70). The measurements are done by visual inspection over the pipes by

boat and approximating the amount of air bubbles coming from each nozzle. A striking feature is the fact

that the first thirty nozzles are performing relatively well, whereas the nozzles after that are clogged much

more. This is due to the fact that the nozzle groups 30 and 63 are gone completely. This yields a big

pressure drop after these nozzle groups, resulting in more clogging in the following nozzles.

20

The production of drinking water from sea water and brackish ground water using sunlight

and nano-particles

Authors: Adam Cahaya (Univ. of Delft), Andrew Croudace (Univ. of Strathclyde, UK) , Giulio D’Odorico

(Univ. of Nijmegen), Nitish Govindarajan (Univ. of Amsterdam), Magdalena Marszalek (Univ. of

Amsterdam), Erik Garnett (Amolf, Amsterdam), Giuseppe Portale (Univ. of Groningen), Rick Wentinck

(Shell Global Solutions, International)

The availability of drinking water is a serious problem in large parts of the world, affecting the quality of

life of billions of people. The challenge is to develop technology that uses abundant and low-cost, non-

toxic, and preferably biodegradable or recyclable materials. In addition, the production process should

be easily scalable and based on a modular design, requiring a minimum infrastructure, making maximal

use of existing technology, whilst being easy to maintain and clean.

Currently, reverse osmosis is under development to become an environmentally acceptable method to

produce drinking water from sea water. The unit uses solar energy from photovoltaic (PV) panels to

press the salinated fluid through the membrane resulting with potable water on the other side. Based

on the current price of PV electricity, reverse osmosis, in regions of interest, produces drinking water for

prices in the range of 0.6 – 2 $/m3. In general, reverse osmosis is an excellent technology when the

volume of water to be treated is high, while its investment cost may remain too high when relatively low

volumes have to be treated.

Although reverse osmosis works well when the salt concentration in the feed water is not too high, as in

sea water, it may be less attractive when the available water contains more salt and other minerals or

when a pipeline for sea water becomes too expensive. Other significant cost may appear to safely

dispose the concentrated brine back into the sea.

Recently, a novel water production method has been proposed by Zhou et al. [1,2] to evaporate the

salty water with sunlight enhancing the evaporation efficiency through a porous plasmonic absorber.

The evaporation takes place along the walls of cylindrical pores of about 300 nm diameter made in an

Al2O3 sheet produced by anodic oxidation from aluminum of less than 0.5 mm thickness, see Figure 1. At

the sunside of the sheet, aluminum nano-particles (NPs) of about 20 nm have been deposited over a

length of a few microns into the pores. In general, there exists an optimum range of diameter of NPs for

maximal sunlight absorbance.

The production of high quality drinking water from very saline sea water of the Dead Sea is impressive.

Another interesting, but practically less relevant, observation is that the evaporation efficiency

remarkably increases for larger light intensities (> 2 sun illumination).

For the FOM workshop of Physics with Industry 2016, Shell proposed to review this new technology and

to investigate whether an integration with waste heat or geothermal heat to boost drinking water

production would be practical.

21

Key processes

Firstly, we identified the key processes. These are: 1) the plasmon-mediated absorption of the sunlight,

2) the evaporation process and 3) the transport of the salt and other minerals.

Plasmon-mediated sunlight absorption

To obtain an efficient adsorption over a broad light spectrum, the plasmons generated by poly-dispersed

NP’s on the surface of the cylindrical pores are essential, in agreement with the claims of the article.

Two main factors are used to broad the absorption spectra and absorbs light more efficiently. At firstly,

an Al2O3 surface layer is responsible for a significant red shift of the absorption. The second and more

important effect is the close packing arrangement of the NPs. The adsorption spectra of these plasmons

significantly broadens when particles are in close contact. There exists an optimum number of particles

in close contact for maximum absorbance. Our calculations show that a similar adsorption as claimed in

the article, and actually even more efficient, can be realised, is a dense packing of nanoparticles with a

wide size distribution in the range 10 – 50 nm radius still maintaining connectivity between them.

Evaporation process

The capillary forces in the aluminum cylindrical pores promote water transport to the sunside of the

porous sheet. It was difficult to estimate the thickness of the water film on the nano-particles, given that

the pores have different diameters. We expect that hydrophilic pores can be beneficial, which would

create a continuous water film for salt transport from the pores back into the bulk fluid but it was

difficult to quantify this effect. We estimate from the heat conductivity properties of the sheet and the

heat fluxes that temperature gradients in the pores are negligible.

Transport of the salt and other minerals through a porous matrix

Thirdly, we investigated possible mass transport limitations for salt to diffuse from the sunside of the

sheet, where the evaporation takes place, to the bulk fluid at the backside of the sheet. For sheets

thinner than about 1 mm we expect no transport limitations for non-concentrated sunlight irradiation.

With an understanding of the key processes, we concluded that the NPs are essential for effective heat

transfer. Ideally, they should be on the sunside of the sheet in the form of a poly-dispersed closely

packed mixture of 10-50 nm radius. The diameter of the pores and salt transport limitations seem less

important.

With this in mind and the aforementioned requirements for drinking water production, we focused on

cheap reactor concepts. From the reverse osmosis process, targeting for 0.6 $/m3 drinking water, and

assuming a 20-year system lifetime and a drinking water production with this technology of 5 l/m2/day

(assuming 5 hours of sunlight per day) we roughly estimated that the total costs per square meter

should not exceed 20 $/m2.

Targeting for such low costs, we concluded that waste heat should already be available in some other

form, aside from the expensive drilling of dedicated geothermal wells.

22

Making maximal use of gravitational flows to avoid expensive pumps, the major costs are in the glass or

polycarbonate cover, which are in the range 2 - 10 $/m2, the porous medium with the NPs and some

vessels to contain the fluids.

For the plasmon-containing nano-porous medium, several cheap options could be exist. Micro-porous

polymer-based matrixes in forms of plastic porous films could be used to support high loaded plasmonic

NPs.

Reactor concepts

We have looked to two essentially different reactor concepts. In the first one, the NP sheet is fully

submerged in fluid and the water boils at 100 C. The advantage of this concept, using boiling water, is

that it automatically kills bacterial life and micro-organisms.

In the second concept, the NP sheet is wetted by capillary forces. This concept can also be easily

integrated in green houses to raise the air humidity around the growing plants. The sheets can be tilted

or simply float in a pond. We anticipate that the very salty brine left after evaporation and collected (in

bins) is dried in the sun at another appropriate location. The two reactor concepts are shown in Figure 2.

An advantage of polycarbonate is that it can be shaped in many different forms using extrusion

technology but it may not have a sufficient lifetime due to aging under continuous sun illumination,

especially in some warm regions of the Earth. When using glass, it should be standard flat glass. High

quality glass, such as used for PV panels, is prohibitively expensive.

To minimise potential toxicity problems and to avoid the growth of micro-organisms and unwanted

bacteria and to simplify the generation of NPs, we believe that silver is the best material for the NPs. We

estimated that the total costs of silver NPs are 2 - 3 $/m2. Of course, aluminum is cheaper but aluminum

NPs are potentially toxic, do not prevent the growth of micro-organisms and have a more expensive

production. We anticipate that the silver containing sheets can be easily recycled to retrieve the silver

NPs.

Potential porous polymer-based sheets with a good control over porosity could be produced via

electrospinning method. These porous sheets are expected to be cheap, i.e. cost well below 1 $/m2.

Hydrophilic robust and mechanically stable polymers should be selected. An interestingly cheap and

environmentally friendly alternative material that we studied, is electrospun cellulose nano-fibre from

cotton waste. Electrospinning of cotton waste has already be proven to be feasible in literature. The

costs for the deposition of silver NP's from a solution on these materials will be minor compared to the

other costs.

To minimise potential toxicity problems, to limit the growth of micro-organisms and unwanted bacteria,

and to simplify the generation of NPs, we believe that silver is the best material for the NPs. We

estimate that the total costs of silver NPs are 2 - 3 $/m2. Of course, aluminum is cheaper but aluminum

23

References

1. L. Zhou et al., Sci. Adv. vol. 2, no. 4, e1501227 (2016)

2. L.Zhou et al., Nature photonics 10, 393-398 (2016)

Figures

24

Figure 1.Porous plasmonic absorber made by aluminum nano-particles in a nano-porous Al2O3 sheet

produced by anodic oxidation, according to Zhou et al. [1,2]. According to these images, the nano-

particles are only deposited over a length of a few microns at the sunside surface of the panel.

25

Figure 2. Reactor concepts making use of micro-porous fibre sheet coated with nano-particles. The

reactor shown in the bottom is integrated in a green house.

The sheet is wetted by capillary forces. The salt flows to the bottom vessel. The sheet is for example

made from electrospun sulfonated polystyrene, cheap, hydrophilic, good control over porosity, robust

and mechanically stable. We choose for plasmonic nano-particles of silver, which has a broader

plasmonic spectrum in the visible range and a high thermal conductivity. In addition, it is affordable,

antibacterial and antimicrobial, robust, non-toxic and easy to deposit by impregnation.

26

1

Microwave Measurement of the Moisture Content in Grains

Do van Lam, Martin Cardarola, Fabiola Gutierrez, Jeff Wood, Andrii Rudavskyi, Maarten Soudijn, Rob Lammertink and Max Hilhorst

Introduction Moisture content in grain is the crucial/price determining parameter of the quality of grain. Typically accepted moisture fraction is within the range of 12% to 17% [1-2]. At higher moisture levels, drying bioprocesses get activated and excessively moist grain is vulnerable towards fungi [3]. Moisture content also strongly can influence the degree of insect attack. Therefore, control and measurement of moisture is a central problem within the grain industry. In order to minimize loss of product (additional cost), it is desirable to be able to assess moisture content prior to silo loading in a simple, economic manner. Over the years, various methods to measure moisture content have been developed [4-6]. Thermogravimetric measurements, where the moisture content is determined via measuring the sample mass before and after oven drying, is the most basic and the most accurate method so far Therefore, this technique is used as a reference for calibration of all the other alternative methods. However, this is a time consuming approach, as well as destructive for the sample. This is also an offline technique, requiring an infrastructure at the loading/unloading site in order to accommodate measurements. The requirement for any new method is: 1) independent of biomass fluctuations or designed in a way to minimize fluctuations in the measurement signal due to non-moisture factors 2) it should be fast, preferably applicable in conditions of the online mode during unloading of the truck, 3) cheap and relatively simple to operate on-site and 4) it should be non-destructive. The last criterion is less crucial, depending on the sampling rate and material usage/measurement given the large amount of biomass offloaded per truck into a silo. For online measurements, electromagnetic probing techniques such as dielectric/microwave or infrared measurements have been employed successfully [7,8]. Furthermore, it has been shown that microwave sensing can work with a simple setup made of cheap components [9]. These methods rely on measuring the dielectric properties of the grain medium as an indicator for the moisture content [10]. However, such approaches require direct control over and probing of the biomass structure, as the dielectric constant or real and imaginary impedances are also a function of the background solids content. If the solids content fluctuates significantly, then either an additional measurement technique is required to decouple moisture from solids content or the measurement device must be designed in such a manner to consistently fill a sample chamber. In our approach, we have investigated the feasibility of static and dynamic online microwave methods to measure moisture content based on simultaneous measurements of microwave propagation through the grain medium employing multiple, but at least two, frequencies. In theory, such an approach allows for simultaneous determination of moisture and solid content [11]. Physical Principle and Setup The dielectric properties of grain are directly related to its moisture content. The highly polar water molecule once embedded into the grain tissue modifies the dielectric constant of the grain, which is at the core of all the microwave-detection schemes. Such schemes usually measure the transmitted power and phase, which can be related to the dielectric properties [7,10]. Due to the ease of

27

2

implementation and low costs involved we choose this scheme for the moisture content measurements [9]. The device concept is represented in Figure 2 (a), where a microwave source is connected to an antenna that acts as a transmitter and a second identical antenna acts as a receiver. The sample is placed in the middle and thus will affect the transmitted power through the device. Figure 2 (b) shows a picture of the prototype, where the white cylinder with red lid is the sample holder that contains the sample (grains) and the metallic cylinder has the transmitter and receiver antennas attached in the rectangular boxes. For the proof-of-principle measurements reported here we used a vector network analyser as a generator and detector of microwaves.

(a)

(b)

Figure 1: Prototype of the measurement device. (a) Setup scheme. A microwave generator is connected to an antenna on one side and on the opposite side an equivalent antenna is placed as a receiver. The transmitted waves will be affected by the presence of any dielectric in the sample holder that will lead to a change in the transmission spectra. (b) Picture of the prototype that was used for the measurements presented in this work.

Results and Discussion From a practical point of view, what should be considered is that a sensor needs a signal that is sensitive to the desired parameter (that what needs to be measured). In order to make sense of the output, a calibration curve is needed that can translate the measured value of the signal into the desired quantity. In our case the signal is the transmission coefficient of the microwave signal through the measuring device containing the grain and the desired parameter is the moisture content of the grain inside. In order to get the calibration curve, we took four reference samples with known moisture content and performed different measurements to see if our signal was sensitive enough to distinguish the samples. Firstly, we conducted static experiments since they are easy to reproduce and manageable with our current device. We measured four samples with moisture content between 13.2% and 15.8%, performing a frequency microwave sweep from 1 to 6 GHz. The transmission can be separated in its absolute value and a phase, that are shown in figure (2) (a) and (b), respectively. Note that the absolute value curve shows clear differences for different moisture content, which is the basis of the proposed sensing method. The phase, on the other hand shows a similar behaviour for all the samples. Different resonance peaks are seen in the spectra and the higher sensitivity is expected at the higher peaks. Thus, we selected the zone with higher signal between 1.3 and 1.8 GHz and for calibration we selected the 3 main peaks present in this range. The spectra in this range are shown in Figure 2 (c) where the vertical lines indicate the selected frequencies for the calibration. We took the transmission

Trasnmitter Receiver

Dielectric

(GHz)

28

3

values for each frequency (normalized to the value at the minimum moisture content) and plotted them as a function of moisture content, as shown in Figure 2 (d). The obtained sensitivities for each frequency are condensed in Table 1.

Frequency [GHz]

Sensitivity [1/% moisture content]

1.49 -0.097 1.55 -0.121 1.62 -0.133

Table 1: Calibration sensitivity obtained for different frequencies In our experimental conditions, such sensitivities allow a precision in the determination of the moisture content of ~0.05%. If we take into account the fluctuations in determining the transmission coefficient for different measurements on the same sample, the precision drops to 1%. This is still quite high compared with established methods but it could be improved with systematic studies on how the grain is packed.

(a)

(b)

(c)

(d)

Figure 2: Proof of principle measurements for the static study. Microwave transmission absolute value (a) and phase (b) for a frequency sweep from 1 to 6 GHz for the four reference samples. A clear difference in absolute transmission can be seen for the different samples with different moisture content. The differences are most pronounced in the range 1.3 to 1.8 GHz, as shown in (c). The local resonance peaks in the spectra were used to determine three convenient working frequencies to use for a calibration curve as shown in (d). The legends in (a)-(c) indicate the moisture content of each sample as determined from thermogravimetric measurements.

29

4

Secondly, we performed some dynamic measurements i.e. with the grain in movement while the measurement is done. To achieve this, we placed the device vertically and let the grain free fall through, as shown in the picture of Figure 3 (a). This emulates the dynamic situation where the grain is being loaded into the silo. The measurement was performed at a fixed frequency of 1.62GHz and as a function of time, starting before the launch – or drop - of the sample. Figure 3 (b) shows the transmittance (T) traces as a function of time, measured for several trials using grain with 15.8% water content and as reference also for the empty and completely filled sample holder. The measurements show high variations from one try to the other, which we attribute to the uncontrolled fluctuations in solid density passing the detection device. These fluctuations are intrinsic to this type of measurement. Therefore, we conclude that this dynamical approach is unsuited for application in its current state. This type of measurement requires an additional independent method to determine the solid content as a function of time to reliably extract the moisture content of the sample.

(a)

(b)

Figure 3: Dynamic measurements. (a) Picture of the experimental setup. (b) Seven dynamic measurements on the same sample (15.8% moisture content) showing different signal patterns that are attributed to different solid content or grain density in the flow.

Given the motivation of easy and fast online measurements and the results presented, we propose a “fill-cell” design for doing quasi-dynamic online measurements. In this approach, samples would be taken by diverting a portion of the flow through a side port with a gating valve, as shown schematically in figure 4. Once the fill-cell is full, a static measurement is done after which the measured solids are released back into the main flow path. The procedure can be repeated several times while the grain is loaded into the silo. While this approach is not fully dynamic, it does potentially allow for online time-sampling of moisture content in a fast, cheap, reliable and non-destructive manner.

g

sample

30

5

Figure 4: Proposed measurement scheme. We propose an online method that uses an auxiliary pipe with a “fill-cell” which is filled up with grain to a certain solid content, performs a static measurement and releases its sample back into the flow. This measurement can be done multiple times while the truck is unloading.

Conclusions We presented a study on the feasibility of using microwaves to determine moisture content while unloading a grain truck. We showed that a static measurement is feasible for detecting the moisture content in grains with the sensitivity and reliability required, provided a precise and detailed calibration is done. Additionally, we tested the measurement principle in a dynamic configuration and we conclude that this approach is not feasible unless a complementary technique is used for solid content determination, but even then such a measurement would be challenging. Finally, we proposed a possible solution to achieve a static measurement in an online fashion that can be implemented for use during loading or unloading of a truck. References [1] D. R. Wilkin and B. C. Stenning Moisture content of cereal grains. AHBD Cereals and Oilseeds. (1989) [2] R. Katz, N. D Collens and A. B. Cardwell Hardness and moisture content of wheat kernels, Robert Katz Publications, 38 (1961) 364-368 [3] CM Christensen, H. H. Kaufmann. Deterioration of Stored Grains by Fungi Ann. Rev Phytopathology. 3(1965) 396 [4] S.W. Pixton and S. Warburton. Moisture content/relative humidity equilibrium of some cereal grains at different temperatures, J. Stored Products. 6 (1971) 283-293. [5] W. Meyer and W. Schilz, A microwave method for density independent determination of the moisture content of solids. Journal of Physics D: Applied Physics 13(1980) [6] A. W. Coats and J. P. Redfern Thermogravimetric analysis. A review, Analyst, 88 (1963) 906-924

31

6

[7] S. Trabelsi, A. Kraszewski, and S. O. Nelson, "Unified calibration method for nondestructive dielectric sensing of moisture content in granular materials," Electronics Letters, vol. 35, pp. 1346-1347, 1999. [8] P. Williams and D. Sobering, “Comparison of commercial near infrared transmittance and reflectance instruments for analysis of whole grains and seeds”, J. Near Infrared Spectrosc. 1, 25–32 (1993). [9] A. Kraszewski, Microwave Aquametry . Piscataway, NJ: IEEE Press, 1996. [10] S. Trabelsi, S. O. Nelson, "Inexpensive microwave moisture sensor for granular materials", in IEEE Antennas and Propagation Society International Symposium, pp. 297-300, 2007. [11] S. Trabelsi, A. Kraszewski, and S. O. Nelson, "Simultaneous determination of density and water content of particulate materials by microwave sensors," Electronics Letters, vol. 33, pp. 874-876, 1997.

32

An energy-efficient, robust and durable poker vibrator for concrete compaction M. Farzanehpour (Free University Amsterdam), R. Gaudenzi (Technical University Delft), K. Haver (University of Amsterdam), A. Moradi (Leiden University), T. van Olst (Lievers Holland BV), R. Sprik (University of Amsterdam). Introduction The construction sector is one of the greatest contributors on the economy and has had a crucial impact in the development of all aspects of human life. From thousands of years, a major role in construction has been played by this composite material we call concrete: a mix of coarse aggregate bonded by a fluid cement that hardens over time. Over the years, continuous effort has been put into improving its structural characteristics (hardness, mass density, etc.) by acting on its chemical composition and deposition methods. One of these methods, called compaction, prescribes the removal of air pockets from the freshly-poured concrete. Nowadays, the most widespread way of implementing compaction consists of inserting a poker vibrator into the poured concrete [1 - 4]. A poker vibrator, or vibrator bottle, is a metallic cylindrical object – from few tens of centimeters to a meter in length – that vibrates owing to the presence of rotating eccentric. This basic technology is developed, proven and so ubiquitous that is used to process the nearly five billion cubic meters of concrete poured annually in the world. However, there are few issues regarding primarily its efficiency, user comfort and durability that we briefly report here:

i. Efficiency. The average energy consumption of a vibrator is between 300 W and 1 kW, with an

engine efficiency of about 60%. Much energy is converted into heat rather than vibrations;

ii. User comfort. The current vibration bottles are relatively heavy due to the steel jacket and

internal motor. Weight is a factor especially when longer periods must be worked with. A similar argument applies to the transmission of vibrations to the hand / arm of the user;

iii. Durability. A poker vibrator suffers from internal and external wear-and-tear due to vibrations

in combination with mechanically and chemically rough materials – concrete (gravel, sand, etc.) and armors steel. Ball bearings supporting the eccentric suffers vibrations and heat and burning of the stator can occur when in overload and with insufficient cooling. This renders the life of the average machine relatively short.

In this work, we have addressed the listed limitations and improved on them by following two strategies: I) working incrementally on the basic design; II) generating an alternative, out-of-the-box concept with analogous functionalities. In the first section we describe improvements of the existing design by investigating the possibilities of adding “smart” electronics to the vibrator. In the second section proposes a solution by improving the mechanics of the existing design while the third sketches a radically new design.

Improvements of existing product

Our results based on experiments with the poker in (wet) sand proved to be successful in identifying the effect of temperature and vibrations on the lifetime and ease of operation of the vibrator. There

33

are currently plans to implement the generated ideas in the existing products of Lievers BV. Detailed results will not be reported here to avoid premature disclosure during the development stage (for further information contact Lievers BV [1]).

Improved eccentric design

While the current poker system is widely used, its design is more than 50 years old and exists mainly of components that were already available at that time. In this section, we propose an improvement of this design that utilizes the same working principle while providing a number of major improvements. Integrated eccentric rotor. The current system uses a squirrel cage rotor, in which a rotating magnetic field is generated through an inner rotor. The motion between this field and the rotor induces electric current in conductive copper bars. These currents react with the magnetic field to produce a force that results in a torque that turns the shaft. The shaft is connected an eccentric weight, which takes up around one thirds of the length of the entire poker.

Figure 4. Integrated eccentric rotor design. The eccentric rotating mass is indicated in green.

An alternative to this construction would be to integrate the eccentric mass into the squirrel cage rotor, so that the rotations do not have to be transferred onto a separate part. We argue that the best way to construct such an integrated mass would be to invert the squirrel cage rotor, letting the outer rotor rotate around an inner stator. This way, the eccentric rotating mass can be attached to the outer rotor, generating vibrations along the entire length of the poker.

Figure 5. Comparison with standard design. Comparison of the current poker (left) and the improved design (right). The green parts indicate the eccentric weights and the red parts indicate the bearings.

Comparing the two systems. In order to ensure generation of similar vibration, we estimated the amplitude of the current system. We first determined the mass of the eccentric part to be 272g and

34

the radius to be 1.7 cm. Then we determined the amplitude of the current poker vibrator using equation (1) and plugging in the mass of the eccentric part.

Circular electromagnetic actuator The design strategy adopted so far is of the incremental type: changes are made on the existing paradigm in order to improve it. While some of the limitations arising from misuse and uniformity of the vibration generation process are solved in this way – see Feedback system and improved eccentric, respectively –, an out-of-the-box thinking is needed if major issues like wear-and-tear want to be solved at the roots. Actuator’s construction and working principle. The driving question at the start of the rethinking process is whether a poker vibrator with no rotating parts and bearings can be produced. The result obtained is sketched in Figure 6. A four-armed shaft (green) made of soft iron and appropriate copper windings is encased in a similarly shaped stator shell (blue-grey) and kept in the geometrical center of the shell by a spring system. In correspondence of each arm and connected to shell, an array of solenoids – or a single, larger solenoid – is placed. The working principle of the device is analogous to that of an induction motor. As a current is applied to the solenoid in the stator, an opposing current is induced in the copper winding of the arm in close proximity of the solenoid. The induced current generates a magnetic field opposing that of the solenoid. The interaction between the two magnetic fields produces a force collinear to the solenoid’s magnetic field that drives the shaft to the stator. By combining the linear motions along the two axes so produced, the shaft and, with that, its center of mass, can be driven in circle around the center of the stator shell (red circle in Figure 6). In practice, this is obtained by applying 90°-phase-shifted AC-currents to each of the solenoids1. By the exact same mechanism of the eccentric mass, the circular motion of the center of mass about the principal axis of the shell produces the desired vibrations.

Figure 6. Circular electromagnetic actuator. The movable shaft (green) made of soft iron is actuated by a series of solenoids and kept in place by springs connected to the outer shell (grey in CAD drawing and blue in the section).

Feasibility assessment. To judge on the technical viability of the presented design, dynamic parameters like the force and solenoid’s currents need to be estimated fixing dimensions and total mass of the system and final vibration amplitude and frequency to be equal to the standard design. To do that, we mapped the oscillating shaft+spring system to a dumped and driven harmonic oscillator. Considering to drive the oscillator at its resonance frequency and a fixed quality factor,

1 By making a three-armed-shaped shaft the system can be driven by the standard AC three-phase current, where the currents are 120°-dephased.

35

the amplitude of driving force is extracted. Assuming, in addition, the solenoids to be dimensioned according to the size of the standard system and with a standard winding density yields a needed peak current through the coils of about 0.2 A. We judged this value to be technically obtainable and therefore demonstrated the feasibility of the proposed design.

Conclusion Our case during the physics with industry workshop was to come up with new concepts for compacting fresh concrete with the goal to improve durability, efficiency and user comfort. Concrete is one of the most important substances used in construction. Although, there has been many new developments in this field however the financial and technical restrictions limit the applicability of the new technologies. So, in this study despite exploring many alternative and revolutionary concepts we decided to propose new designs which are easy to manufacture and maintain. Our proposal focuses in two areas. First, improving the efficiency and durability of the exiting design. Second, redesigning the poker internal structure to improve the durability and efficiency. Concrete compactors have simple internal construction and as a result have relatively low failure rate. However, since these appliances need to work in harsh environments there are many incidences that their lifetime can be drastically reduced. In this study, we considered three major scenarios that put pressure on the internal construction of a poker, specially, the bearings: 1. Poker stays working for an extended time outside of concrete which makes the poker overheat. 2. The poker is turned on but left unattended on hard a surface. 3. The poker is inside the concrete but in contact with steel reinforcement bars. We developed an approach that can avoid excessive damage to the vibrator without the need to completely modify the existing design and control electronics. In the second part of this study we shifted our focus to redesigning the poker internal structure with the aim to reduce the moving parts which directly reduces the weight and also increases the life time of the system. Another benefit of the new designs is that they will distribute vibrations more evenly along the length of the poker and thus improve the efficiency of the poker. The first redesign is based on eccentric motion with an integrated eccentric into the rotor. Another improvement is the omission of one bearing which is the part with most failure rate in the current system. This design shares the same working principle with the current build which allows it to be manufactured with minimum changes in the production technologies. As mentioned many times the weakest points in the current construction are the bearings. So, in the second design we remove the rotating parts all together and create the vibrations directly using the circular electromagnetic actuator. This makes the internal structure of the poker much simpler and reduces the moving parts drastically. Therefore, we expect this design to deliver much longer lifetime than the current system. Finally, we made an approximate calculation of the parameters for both of the designs with the goal to create the same amplitude of vibrations produced by the current system. We believe that, the current study can be the starting point for the future works on the alternative and more effective poker designs. References [1] See e.g. the product line of Lievers Holland BV: http://www.lieversholland.com/products/poker-vibrators/ [2] IB 46: Vibration of Concrete, Cement & Concrete Association of New Zealand, IB 46, March 2005. [3] P.F.G. Banfill, Rheology of Fresh Cement and Concrete, Rheology Reviews, 61 – 130 (2006).

36

[4] P.F.G. Banfill, M.A.O.M. Teixeira, R.J.M. Craik, Rheology and vibration of fresh concrete: Predicting the radius of action of poker vibrators from wave propagation, Cement and Concrete Research 41, 932-941 (2011).

37

Proceedings Physics with Industry 2016

- 1 -

Philips

Generation of ultrapure droplets using a mesh nebulizer

Ehsan Barati1, Lantian Chang1, Burak Eral2, Pascal de Graaf3, Badr Kaoui4, Marco Marciani5, Naghmeh Mohammadi6, Jeroen Rodenburg6, Maria Sovago7, Alwin Verschueren3, Stephen Wilson8, Yujie Zhou5

1 University of Twente, The Netherlands 2 Delft University of Technology, The Netherlands 3 Philips Research, The Netherlands 4 CNRS - Université de Technologie de Compiègne (BMBI), Compiègne, France 5 Leiden University, The Netherlands 6 Utrecht University, The Netherlands 7 NWO, The Netherlands 8 University of Strathclyde, Glasgow, Scotland, UK

1. Company information Royal Philips is a leading health technology company focused on improving people's health and enabling better outcomes across the health continuum from healthy living and prevention, to diagnosis, treatment and home care. Philips leverages advanced technology and deep clinical and consumer insights to deliver integrated solutions. The company is a leader in diagnostic imaging, image-guided therapy, patient monitoring and health informatics, as well as in consumer health and home care. Headquartered in the Netherlands, Philips' health technology portfolio generated 2015 sales of EUR 16.8 billion and employs approximately 69,000 employees with sales and services in more than 100 countries.

2. Problem 2.1. The background and urgency of the problem Mesh nebulizers are commonly used to atomize liquids into billions of micrometer sized droplets, intended for inhaling medication in an aerosol form into the lungs of patients, for treatment of cystic fibrosis, asthma, COPD and other respiratory diseases. A limitation of mesh nebulizers is that droplet formation is observed to stop working for too low electrical conductivity of the liquid (typically corresponding to < 0.1mM NaCl). As a consequence, mesh nebulizers can currently not generate ultrapure water droplets. Solving this could enable new medical and non-medical applications beyond respiratory drug delivery. 2.2 Why you need help from the physics community The mechanism why mesh nebulizers cannot generate droplets from ultrapure water is poorly understood. It is speculated that low conductivity of the fluid allows electrical charging of the droplets, and resulting electrostatic attraction forces with the mesh prevent the droplets to escape. Philps sought the help of the physics community to build up understanding of the multi-disciplinary processes that may contribute to understanding of

38

Proceedings Physics with Industry 2016

- 2 -

this phenomena, to ultimately guide towards solution directions to overcome the problem. 2.3 Which physics disciplines you expect to be relevant to solve this problem We believe that this problem can only be tackled by combining the insights of multiple physics disciplines: fluid mechanics, acoustics, microfluidics, interface and colloid science, electrokinetics, electrostatics, and perhaps more. 2.4 Possible solutions or directions toward solutions to the problem To build up understanding on the fundamental aspects of the problem, it may be helpful to perform calculations or simulations on the relevant time scales of the droplet formation, its key parameters, the dynamics of ions at low concentrations, electrostatic forces between droplets and mesh and relative to their escape velocity. Possible solution directions to create ultrapure droplets may be found in exploiting additional electrical potentials on the device, optimizing the driving frequency of the nebulizer, adding electrical discharge periods and adapting mesh material properties. The feasibility of these approaches may be estimated using first principle physics based on key parameters. 2.5 Boundary conditions (e.g. technical, organisational, or budgetary requirements) The ultimate goal to create ultrapure droplets should be achieved by modification of existing mesh nebulizers, and ideally should not increase the form factor, complexity and cost of the device significantly.

3. Introduction and problem statement

3.1 Motivation Mesh nebulizers are commonly used to atomize liquids into billions of micrometer sized droplets, intended for inhaling medication in an aerosol form into the lungs of patients, for treatment of cystic fibrosis, asthma, COPD and other respiratory diseases. In addition to drug delivery, if a mesh nebulizer could be used to atomize pure water, this could enable new medical and non-medical applications beyond respiratory drug delivery. However, as will be discussed below, it is observed that mesh nebulizers are unable to atomize pure water. The “Physics with Industry” challenge is to identify the underlying physical mechanism responsible for this limitation.

3.2 Mesh Nebulizers The challenge revolves around two types of mesh nebulizers already available in the market: with either an active or a passive mesh [1].

In active mesh devices a dome shaped circular mesh is located at the bottom of a liquid reservoir. At rest conditions, the small dimensions of the mesh nozzles prevent liquid to pass through the mesh, because of the water surface tension. The edge of the mesh is in contact with a piezo-electric material, whose electric-induced vibration leads the mesh to stretch and compress. Due to the curvature of the mesh surface, its center is led to move up and down inducing the water at rest to be atomized through the nozzles, thereby forming droplets. An example of an active mesh device is shown in Fig. 1.

39

Proceedings Physics with Industry 2016

- 3 -

Fig. 1: Schematic illustration of an active mesh nebulizer (left), with details of the piezo and the mesh (center) and of the nozzle (right). Adapted from Ref. [2].

Active meshes have a simple geometry. However, they have a limited lifetime resulting from mechanical deformation and stress. To improve on this shortcoming, passive mesh devices have been developed. The domed mesh is replaced by flat stiff one, and the piezo-electric actuator is only in contact with the liquid not directly with the mesh itself. The droplet formation is induced by the vibration of the piezo: acoustic pressure waves force the liquid though the nozzles in the form of droplets. An example of such a device is shown in Fig. 2.

Fig. 2: Schematic illustration of a passive mesh nebulizer (left), with details of the mesh (right). Adapted from Ref. [2].

Although the two devices are different, their working principle is essentially the same. Active meshes produce droplets the same way passive ones do, if one considers the dynamics from a reference frame of the mesh nozzle; water and mesh are moving relative to each other. For the clarity of explanation, in the following we will assume always the water to be in movement and the mesh to be at rest. As an indication of the droplet formation process, in Fig. 3 numerical modelling results [3] are shown. Here one full cycle (lasting 1 microsecond) is shown of periodic pressure variation, showing how the fluid by periodic pressure is pushed through the nozzles, and by surface tension and inertia forms droplets.

40

Proceedings Physics with Industry 2016

- 4 -

Fig. 3: Numerical modelling results of droplet formation in a single nozzle of the mesh. Here one full cycle (lasting 1 microsecond) is shown of periodic pressure variation. From [3].

4. The challenge

4.1 Problem statement

Fig. 4: Water output rate of a nebulizer as a function of NaCl concentration. Notice the logarithmic scale on the horizontal axis. Adapted from Ref. [4].

Performance of mesh nebulizers with various liquids has been studied by some authors [4,5]. It is of our interest to consider water at different salt (e.g. NaCl) concentrations. Testing on an active mesh nebulizer, M. Beck-Broichsitter and collaborators observed a clear decrease of the output rate as the NaCl concentration gets below 0.1mM, see Fig 4. With “output rate decrease” meaning essentially that a lower number of droplets is produced by the device per unit time. From our own experimentation we have reproduced this phenomenon using both active and passive mesh nebulizers, both with metallic (nickel-palladium) and polymer (polyimide) mesh materials. Repeating the experiment with many identical devices, we have always observed a marked decreased output rate at vanishing NaCl concentration.

Our challenge, therefore, is to answer this question: What could be the physical mechanism that causes the failure in the droplet formation at low salt levels?

4.2 Preliminary considerations The physics of droplet formation is clearly a matter of hydrodynamics. Having in mind the plot in Fig. 4 and, at first, the Navier-Stokes equation for electrically neutral Newtonian

41

Proceedings Physics with Industry 2016

- 5 -

liquids, one should ask himself what are the hydrodynamical parameters that could bear relevant changes when the salt concentration (C) is varied in the same interval of the plot. We list now the degree of change of the main parameters as C is varied from 10-2 mol/m3 to 5

mol/m3. We found that (i) the viscosity has a 2% increase [2]; (ii) the surface (water/air) tension has a 0.7% decrease [2]; (iii) the density has a 0.02% increase [6]. Thus, changes are very small, unlikely to cause the dramatic effect on droplet formation.

An indication that the elektrokinetics must not be disregarded comes from realizing that conductivity has a linear dependence on the salt concentration. When C is varied from 10-2

mol/m3 to 10-1 mol/m3, it has a 1000% increase [7].

This fact led us to consider the typical phenomena associated to charged fluids such as interfaces and droplet charging, screening mechanisms and electrowetting. All these phenomena are presented and discussed in the following sections.

4.3 Flow electrification induced charging In this section, we investigate the charging effect due to flow electrification. We first briefly describe the phenomenon of flow electrification. Then we calculate the net charge density flows in the nozzle. At the end of this section, we propose a few suggestions for further research.

The understanding of flow electrification is important for many industrial applications. In low conductivity liquid, a voltage can build up to a very high level that cause electrical discharges thus lead to electrostatic hazards [8]. The DI water we used in our experiments has low conductivity, which could have a large charge or voltage accumulation and prevent the droplet formation or flyaway from the nozzle.

The charge distribution at the nozzle cross section is indicated in Fig. 5. When a liquid is in contact with a solid, a physicochemical reaction leads to a electrical double layer [8,9]. The charge density decreases from the boundary to the center of the nozzle.

Fig. 5: Charge distribution at the nozzle vertical cross section. The dashed vertical line represents the center of a cylindrical nozzle. And r is the radius and ρw is the charge density of the liquid at the boundary.

The charge density distribuation also depends on the geometry of the nozzle. For our circular cross section the charge density distribuation is [8]

0 0

0 0

( ) ,

w

I rr

I R (1)

42

Proceedings Physics with Industry 2016

- 6 -

where r is the radial position (0 ≤ r ≤ R), R is the radius of the nozzle (1.25 μm), ρw is the charge density of the liquid at the boundary, I0 is the Bessel modified function of zero order and δ0 is the Debye length (diffuse layer thickness). The Debye length can be calculated with

B0 2 ,

2 A

k T

N e C (2)

where ε is the permittivity of the liquid, kB is the Boltzmann constant, T is the absolute temperature in kelvins, NA is the Avogadro number, e is the elementary charge and C is the ionic strength (ion density) and the unit is mol/m3.

The calculated normalized charge density, ( ) wr , at different ionic is shown in Fig. 6. The

ideal ID water (with pH=7) has an ionic strength of 10-4 mol/m3, which results in a larger normalized charge density profile than at higher ionic strength cases.

Fig. 6: Normalized charge density distribuation at different ionic strength cases. Normalized to their own charge density at the boundary (ρwn). The dashed black line shows a fully developed laminar flow speed profile, which is normailized to the speed at the center of the cylinder.

The charge density multiplied with the normalized laminar flow speed profile, integrated over the radius of the nozzle and divided by the area of the nozzle horizontal cross section, results in the averaged charge density moving with the flow. The corresponding results are shown in Table 1.

Table 1: Average moving charge density.

Ionic strength (mol/m3) 10-4 10-3 10-2 1

Average moving charge density qa (C/m3)

0.79 ρw1 0.27 ρw2 4×10-2 ρw3 5×10-4 ρw4

In this calculation, the charge density at the boundary of each individual case remains unknow. Therefore, one of the follow up studies should find out the ρw in every case. It can

43

Proceedings Physics with Industry 2016

- 7 -

be a function of the material of the solid boundary and can also change with the ion density of the liquid. After knowing the charge density at the boundary, we could quantitatively know the amount of charge in each droplet and the electrical force between the droplet and the remainder of the system.

All the calculations discussed in this section are based on constant liquid flow in the nozzle. However, our real system is working in an oscillating manner. Thus, the next step should be investigation of the dynamics of our system. This can include but not limited to study the charge in the first droplet and the charge density in an oscillating liquid.

5. Potential failure modes due to charge accumulation To fully understand aerosol droplets formation (aerosolization) process one needs to consider, as we do in this manuscript, different probable factors which influence this process e.g., charge-induced wetting, Coulomb repulsion between two droplets just after formation on the nozzle head as well as stabilization of droplet neck.

5.1 Charge-induced wetting The charge-induced wetting resembles the more famous phenomenon called electrowetting during which electric charge, supplied by an external electric filed, changes liquids’ properties e.g., wettability during droplet formation. As its title suggests the charge-induced wetting is closely related to the wettability of charged substances and it has been previously employed for investigation of charged-aerosols behaviour e.g. for the charged-aerosol-related cloud formation [10]. One of the main parameters entering the wetting phenomenon is the so-called contact angle which illustrates the curvature of a liquid droplet on a surface and can be calculated using the Young-Dupre equation which relates the contact angle to the so-called spreading parameter S as

1 ,LGS cos (3)

where LG is the liquid-gas phase surface tension. The small contact angle corresponds to

high wettability for which the liquid more spreads on the surface and vice versa. It is worth mentioning that the wettability is directly related to the surface tension of the liquid as different contact angles correspond to different surface tensions. Though it is possible to determine the contact angle for a specific system, we herein provide a quantitative description of the contact angle in charge-induced wetting rather than its direct calculations for the investigated systems.

neutral liquid (no induced charge) charge-induced liquid

Fig. 7: A schematic picture on the modification of contact angle due to charge-induced wetting during droplet formation.

44

Proceedings Physics with Industry 2016

- 8 -

In electrowetting the created field tends to pull the droplet down onto the electrode lowering the contact angle. In a very similar way, the contact angle is modified in charge-induced wetting. The modification of the contact angles in the investigated systems is attributed to the different charge densities in the systems (see Fig. 7). The electrostatic interaction between the charged droplet and the surface pulls the droplet down and favours the wettability of the surface. This can be mathematically deduced from the Young-Dupre equation. For a fixed, liquid-gas phase surface tension, spreading parameter increases with decreasing the contact angle and thus favours the wttability of the surface. This coincides with our experiment for pure water in which droplets spread on the nozzle head and fail to get formed. Note that, charged-induced wetting may play a role, with spreading the liquid onto the mesh surface surrounding the nozzle, and thereby prevent droplet formation if a metallic mesh is used. However, it is revealed in experiment that droplet formation at low ionic strengths is prevented also in non-metallic meshes. This leads us to the concnlusion that charge-induced wetting is not the mechanism of our interest here and it is considered as one of the failure modes to droplet formation.

5.2. Neck stabilization In the process of droplet formation, the water leaving the nozzle needs to disconnect from the water in the reservoir. The presence of net charge in the droplet can prevent this disconnection process known as ‘pinching off’; this section explains how.

For this failure mode, it is important to distinguish two different physical mechanisms that can create droplets in the first place. Both processes correspond to different regimes the nebulizer can, at least theoretically, operate in, as they correspond to different driving frequencies (and possibly also different peak pressures). In the first regime, droplet formation is a result of the oscillating water pressure in the nozzle. Droplet formation in this regime proceeds as sketched in Fig. 8: at the time of high water pressure in the nozzle, water builds upward inertia and is expelled from the nozzle. Half a period of the pressure oscillation later, this upward inertia of the top of the water column is still present, yet the low pressure at the nozzle causes water to retract inwards. These two opposing motions lead to rupture of the water formation and to escape of the water droplet. In the second regime, the column of water expelled during one oscillation is so long that it is unstable to perturbations in its (vertical) surface. Consequently, the water formation breaks up as a result of growing perturbations. This mechanism, known as the Rayleigh instability [11], is expected to be responsible for the pinching off in the regime of low driving frequency, and is the mechanism used in inkjet printers [12]. To determine which regime the nebulizer operates in, we estimate relevant time and length scales. In order for the second regime to be important, the perturbations need to be able to grow within one oscillation period. The

timescale for these perturbations to grow is 303 0.5h st r g m= » for the water density

-31000 kgm , water column radius 0 1.25 mh (coinciding with the nozzle radius), and

air-water surface tension 10.07 Nm [12]. This timescale t is of the same order as the

period of the pressure oscillations T 1μs. A second condition is that the length of the water column (or ‘jet’) emitted during one oscillation must be larger than the circumference of the nozzle for the jet to be unstable (when approximated to be cylindrically shaped). An upper bound for the length L of the jet is found as L≲vT=10 μm, where v=10 m/s is the typical velocity of an emitted droplet. This upper bound is of the same order as the nozzle circumference 2πh0 8μm. However, we expect realistic values for L to be somewhat smaller. In addition, for a length L 2πh0, the only unstable perturbations take more time thant to grow. For these reasons, we think the nebulizer operates in the first regime. However, this

45

Proceedings Physics with Industry 2016

- 9 -

simple observation shows that the nebulizer is actually close to operating in the second regime.

5.2.1. Neck stabilization in the operating regime We now discuss how the presence of net charge can prevent the droplet from pinching off in the operating (i.e. the first) regime. At the latest stage of droplet formation, the droplet is connected to the remaining liquid by only a think neck, as illustrated in Fig. 9, which shows a snapshot of a simulation of this stage of the droplet formation. For simplicity, we approximate this neck by a thin cylinder of radius R=100 nm. For a successful droplet formation, the reason that this neck thins even further such that the droplet eventually pinches off, is that the surface tension in the air-water interface provides a force that pulls the surface inwards. The magnitude of this inward force (per unit area) is of the order of the Laplace pressure 1Lp Rg= bar. If the neck obtains a net charge, the repulsive

electrostatic forces the net charges exert on each other provide a counteracting outward force. To estimate the magnitude of this outward pressure, we approximate the neck as a cylinder with all its net charge Ne (where e is the fundamental charge) located at the surface as a surface charge density =Ne/(2 RL), where L is the vertical length of the cylinder. The outward pressure is then approximated as the electrostatic pressure 2

0(2 )e rp s e e= , where

r is the relative permittivity of water and 0 is the vacuum permittivity [3]. The condition

for this electrostatic repulsion to be significant is L ep p , which happens for

1 22 250

2

810r RL

Ne

p ge eæ ö÷ç ÷ »ç ÷ç ÷çè ø (4)

net charges in the neck, where we assumed 1 mL . In future studies, this number can be used as an indicator for the relevance of the neck stabilization failure mode.

5.2.2. The Rayleigh breakup regime Here we investigate the Rayleigh instability. As discussed above, we expect the nebulizer to be close to the regime where this instability is important. Therefore, it is worthwile to

Fig. 8: Droplet formation due to pressure oscillations. At high water pressure in the nozzle, upward inertia (red arrows) builds up and expels water. Half an oscillation further in time, water retracts again, and the still present upward inertia of the droplet-to-be causes the water formation to rupture and to escape.

Fig. 9: Snapshot of a simulation of droplet formation, showing the stage just before the droplet pinches off. The droplet-to-be connects via a long thin neck to the water reservoir.

46

Proceedings Physics with Industry 2016

- 10 -

investigate how droplet formation by the Rayleigh instability is affected by flow electrification. The Rayleigh stability theory states that for a cylindrical liquid jet, any perturbation of the radius of the jet of sufficiently long wavelength will result in the reduction of the surface area and thus enhance the perturbation. The jet radius h subject to the perturbation takes the form [11]

(7)

where 0h is the mean radius, is the coordinate of the jet axis, is the wave number of the

spatio perturbation wave along -direction, is the perturbation amplitude, and is the time frequency of the perturbation (See Fig. 10). We keep our discussion for the lowest azimuthal mode so no azimuthal angle is involved.

Fig. 10: The cylindrical liquid jet. The criterion for stability for perturbation of given wave number is whether the time dependent part of the perturbation is increasing, or in other words, whether the

dimensionless growth rate ( )i kw t- is a real positive number. Here 303 ht r g= is the

associated time of growth, is the surface tension and the liquid density. For neutral jets we have the dispersion relation

(8) The dispersion relation for jets with charges is

(9) where

(10)

and is the surface charge density. To show the effect of the extra term containing due to electrical charges, we plot in Fig. 11 the dimensionless growth rate versus the wave

47

Proceedings Physics with Industry 2016

- 11 -

number 0kh . We see that the charged jet has a wider instability range of wave number and a

higher growth rate. This demonstrates the destabilization effect of the jet electrification which facilitates the droplet formation process.

Fig. 11: The dimensionless growth rate versus the dimensionless sinusoidal perturbation wave number 0kh . The left curve is for the neutral jets and the right curve is for the charged jet with (cf. Ref. [11]). We conclude that flow electrification actually enhances droplet formation by the Rayleigh instability mechanism for deionized water. If it is possible for the nebulizer to operate in an (e.g. lower-frequency) regime where this mechanism becomes important, this effect can possibly solve the practical problem of this workshop.

5.3. Droplet repulsion The other scenario is based on the repulsion of same charged droplets. In this scenario, the released droplets from the nozzle repel each other due to the Coulomb force between their non-zero net charges. As a result of this repulsion, droplets lose their initial velocity and return back to the nozzle. In order to investigate this scenario and its potential role in the problem, we look at a simplified version of this scenario illustrated in Fig. 12.

Fig. 12: A schematic view of two droplets released from the nozzle

48

Proceedings Physics with Industry 2016

- 12 -

In a simplified version, two droplets have just been released from the same nozzle with an initial velocity of 10 m/s. When droplet 2 leaves the nozzle, due to the Coulomb force between the net charges of the two droplets, droplet 1 accelerates and subsequently, droplet 2 decelerates. The Coulomb force on droplet 1, assuming both droplets are positively and

equally charged can be shown according to 2

204

qmx

x . In this equation, m and q are the

mass and the charge of each droplet, respectively and x is the distance between the two droplets.

By solving this differential equation, one can calculate the dependence of the velocity of droplet 1 on the distance between the two droplets according to

22 2

00 0

1 1 1 1( )

2 2 4q

x vm x x

(5)

where 0v is the initial velocity of the droplet and 0x is the initial distance between the two droplets. This equation can be rewritten for droplet 2 according to

22 2

00 0

1 1 1 1( ).

2 2 4q

x vm x x

(6)

In addition, mass of the droplets can be given by 343

m v r , where r is the radius of the

droplet.

Table 2:

parameter Value

0x 10 μm

0v 10 m/s

r 2 μm

q ?

49

Proceedings Physics with Industry 2016

- 13 -

Fig. 13: Velocity of the droplets as a function of the distance between the two droplets released from the nozzle with equal net charge of q=103 e.

Fig. 14: Velocity of the droplets as a function of the distance between the two droplets released from the nozzle with equal net charge of q=104 e.

By plugging the parameters from table 1 into the velocity equations one can investigate the dependence of the velocity on the distance. Figure 13 and 14 show the velocity of the droplets as a function of the distance between the two droplets released from the nozzle. Figure 13 shows that the net charge of q=103 e in each droplet is not enough for the Coulomb force to decelerate droplet 2 shown with the red dotted line. However, if the net charge increases to e.g. q=104 e, the Coulomb force will cause enough deceleration for droplet 2 to fully lose its momentum at around x=30 μm. This effect has been shown in Fig. 14.

In conclusion, if the released droplets from the nozzles have sufficient net charge of the same sign this mechanism can be a potential reason for the droplets to lose their momentum and accumulate on the nozzle, namely “the crying effect”.

50

Proceedings Physics with Industry 2016

- 14 -

6. Experimental suggestions In this section, we will provide a set of experimental suggestions for future work. We believe the suggestions will provide experimental proof to validate or reject the theoretical framework we provided.

6.1. Measuring the streaming potential

All of the suggested failure modes contained a charging mechanism. A promising candidate for charging mechanism is flow electrification also referred to as streaming potential. It is observed when the mobile part of this counter charge is dragged along by the hydrodynamic flow discussed in detail in section 4.3. Therefore we suggest measuring the streaming potential in our system to validate/reject this hypothesis. For this purpose, we suggest to operate the nebulizer in a different way i.e. use it as a porous plug and pump the fluids using hydrostatic pressure. The applied static pressure will drive the fluid across the pores and drag the mobile countercharge. We propose to measure the streaming potential/current using a simple circuit consisting of Ag/AgCl electrodes and a multimeter as suggested by Olthuis et al [14] and others [15-16]. Figure 1 shows the suggested experimental scheme where the pressure differential Ph-Pl drives the fluid through the nebulizer mesh of given diameter while the multimeter measures the streaming potential. Pressure at the first fluid chamber (Ph) is maintained by a pressure gauge at 1 bar and the pressure at the second fluid chamber (Pl) is maintained by a water jet pump at a significantly lower pressure ~ 0.03 bar.

nebulizer mesh

fluid chamber 1

Pl

Ph

fluid chamber 2

electrode

electrode

water jet pump

Rext

2

1

I

1 V mul meter 2 ground

Vs

Fig. 15: Schematic illustration of experimental system measuring the streaming potential created by flowing of liquids through the nebulizer. Adapted from Ref. [14].

The suggested experimental scheme in Fig. 15 can enable direct measurement of streaming potential (Vs). We suggest measuring the streaming potential with electrolyte solutions of monovalent ion KCl at different concentrations (10-3 M to 10-6 M). The proposed experiments will be conducted while the piezo element integrated into the nebulizer mesh is not functioning to focus solely on the flow electrification. Operation of the piezo element can introduce interesting physical phenomenon hence definitely worth exploring. Also the

51

Proceedings Physics with Industry 2016

- 15 -

influence of air bubbles and impurities in the system is worth further exploration as it has been recognized to influence the streaming potential [17].

6.2. Measuring the charge on droplets The experiments suggested in section 6.2 measures an overall streaming potential when the mesh is fully immersed in fluid. However, the mesh is in contact with air during the regular operation of the nebulizer therefore direct measurements of the charge on nebulized droplets can provide useful quantitative information. Furthermore, there is a distinct possibility hinted by the Philips team’s simulation results that only a small fraction of nozzles contribute to formation of droplets. Therefore, direct measurements of the aerosol size are essential for drawing conclusions on the underlying failure modes. The measurement techniques for measuring the charge of droplets in aerosols are abundant in literature. Some of the methods mentioned in the literature are Faraday pail method, conducting probe, electrostatic cascade impactor, bipolar charge measurement system and laser Doppler velocimetry [18]. Simultaneous measurements of charge and size distributions in aerosols in similar range droplet size as our system have been reported in literature using phase Doppler anemometry (PDA) [19]. Kulon et al. used the PDA system to track the motion of charged particles in the presence of a DC electric field. By solving the equation of particle motion in a viscous medium combined with the simultaneous measurement of its size and velocity, the magnitude as well as the polarity of the particle charge can be obtained.

6.3. Imaging the mode of failure Direct imaging of the nebulizer during its operation and how it fails can provide useful information in identifying eliminating the correct failure mode scenario. For this purpose, we suggest imaging at relevant length (single pore at micron scale & complete mesh at millimeter scale) and time (single droplet creation within a period of oscillation i.e. microseconds and moment of failure i.e. order seconds) scales. As the successful droplet release is periodic stereoscopic imaging can provide the opportunity to image at relatively low-frame hence avoid using expensive high frame rate cameras.

To this end, we suggest the following experiment. Multiscale imaging with in-situ control of salt concentration. We propose to decrease the salt concentration gradually, and image the transition from the successful operation into failure. If increasing the salt concentration turns out to be more convenient experimentally, we can also observe the transition from failure to successful operation. In early demo experiments, we observed a drop of saline solution added to the water tank of the nebulizer in failure mode struggling to create droplets with dionized water instantaneously enabled formations of droplets. Therefore we predict the suggested experiment focusing on the transition from failure to successful operation can reveal the underlying physics of this challenge.

7. References [1] Ari A. Jet. “Ultrasonic, and Mesh Nebulizers: An Evaluation of Nebulizers for Better Clinical Outcomes”, Eurasian Journal of Pulmonology 16 (2014) 1-7; T. C. Carvalho and J. T. McConville, “The function and performance of aqueous aerosol devices for inhalation therapy”, J. Pharm. Pharmacol. 68 (2016) 556–578

[2] T. Ghazanfari, A. M. A. Elhissi, Z. Ding, K. M. G. Taylor. “The influence of fluid physicochemical properties on vibrating-mesh nebulization”, Int. J. Pharm. 339 (2007) 103–111

[3] A. R. M. Verschueren et al. “Multiphysics modeling of a medical nebulizer device: the path from piezo to droplet”, Keynote talk at Comsol conference 2013, Rotterdam (2013)

52

Proceedings Physics with Industry 2016

- 16 -

[4] M. Beck-Broichsitter, N. Oesterheld, M.C. Knuedeler, W. Seeger, T. Schmehl. “On the correlation of output rate and aerodynamic characteristics in vibrating-mesh-based aqueous aerosol delivery”, Int. J. Pharm. 461 (2014) 34–37

[5] A. Bohr, M. Beck-Broichsitter. “Generation of tailored aerosols for inhalative drug delivery employing recent vibrating-mesh nebulizer systems”, Vol. 6, No. 5, (2015) 621–636

[6] Water density Calculator, http://www.csgnetwork.com/h2odenscalc.html. The calculator refers to seawater, where many different salts are dissolved. However, notice that sodium chloride has by far the biggest contribution (see https://en.wikipedia.org/wiki/Seawater). A temperature of 23 °C has been set for the alculation

[7] “Conductivity ordering guide”. EXW Foxboro (1999, retrieved 2013)

[8] O. Moreau, et al., “Influence of the wall shearing stress on flow electrification”. 3rd ISNPEDADM, Reunion (2015)

[9] G. Touchard, “Flow electrification of liquids”, J. Electrostat 51 (2001) 440–447

[10] B. Mook Weon and J. Ho Je, “Charge-induced wetting of aerosols”, Appl. Phys. Lett. 96 (2010) 194101

[11] J. Eggers and E. Villermaux, Rep. Prog. Phys. 71 (2008) 036601

[12] Y. Son, Macromolecules 36 (2003) 5825

[13] D. J. Griffiths “Introduction to Electrodynamics”, San Francisco, CA: Pearson Benjamin Cummings. (2008)

[14] W. Olthuis, B. Schippers, J. Eijkel, A. van den Berg Sensors and Actuators B 111–112 (2005) 385–389

[15] C. Werner, R. Zimmermann, T. Kratzmuller, “Streaming potential and streaming current measurements at planar solid/liquid interfaces for simultaneous determination of zeta potential and surface conductivity”, Colloids Surf. 192 (2001) 205–213.

[16] X. Xuan, D. Li, “Analysis of electrokinetic flow in microfluidic networks”, J. Micromech. Microeng. 14 (2004) 290–298.

[17] T. Nguyen, Y. Xie, L. J. de Vreede, A. van den Berg and J. C. T. Eijkel, Lab Chip (2013) 13, 3210-3216

[18] S. H. Kim, K. S. Woo, B. Y. H. Liu, M. R. Zachariah, “Method of measuring charge distribution of nanosized aerosols”, Journal of Colloid and Interface Science 282 (2005) 46–57

[19] J. Kulon, B. E. Malyan, W. Balachandran IEEE Transactions on Industry Applications Vol. 39 Issue: 5 (2003) 1522–1528

53