galbraith society undergraduate engineering journalalice.ye/... · galbraith society club meeting:...

Galbraith Society Undergraduate Engineering Journal

Galbraith Society

September 2015

Contents

About the Journal ii

Directors’ Message iv

Part I: Journal Papers 1

Evaluating Phosphate Levels In Green Roofs Of Different Growing Media 2Allan Pretti Ogura, Jenny Hill and Jennifer Drake

Characterization of Hydrogen Side Reaction for Zinc-Air Fuel Cell 7Hui Huang Hoe and Donald Kirk

Investigation of Dephosphorization of Ferrosilicon Alloy Subjected to Electromagnetic Levitation and Argon-Hydrogen Gas Flow Mixture 11

Andrew Hue, Katherine Le, Wing Yi (Roxana) Pao, Yindong Yang and Alexander McLean

An Evaluation of the Stability of the Boeing 747 Flight Simulator Model using a CN,β Dynamic Map 17Arthur Brown and Peter R. Grant

The Design and Testing of a Student-Built Paraffin-Aluminum-Nitrous Oxide Hybrid Sounding Rocket 21Jeffrey R. Osborne, Ashis Ghosh, Adam De Biasi, Hayden Lau and Jeremy Wang

Feasibility of Single-Link Inverted Pendulum Models for Human Standing Posture 30Kai Lon Fok and Kei Masani

Part II: News Articles 35

Legacy Uranium Mining Threatens Water Supply 36Sharon Mandair

Why Does the Same iPhone Come Out Every Year? 39Sam Osia

The Growing Field of Personal Health Care 40Shendu Ma

About the Journal

About the Club

The Galbraith Society began as an academic society formed by passionate engineering undergraduates. Since 2012, the

Galbraith Society has been focused on improving the academic experience for engineering students from all disciplines.

Carrying on its original tradition, the group provides opportunities for students to gain exposure to their fields of interest

and network with other highly motivated students through its Research Experience (REX) Program, Undergraduate

Engineering Journal, and Symposia/Workshops that appeal to students’ passion for engineering and research.

The GS Undergraduate Engineering Journal (GSUEJ) is a new initiative by the Galbraith Society that makes it possible

for any engineering student or undergraduate researcher at the University of Toronto to publish the academic research

that they conducted. This initiative was introduced this year to showcase students’ achievements in research, while

simultaneously providing students with an opportunity to develop effective communication skills. This program

encourages students to publish research journal articles in their interested area of research, and also teaches them how to

write and edit articles through a series of research writing workshops held during the year.

About the Cover

It seemed only fitting to dedicate the inaugural issue of the Galbraith Society Undergraduate Engineering Journal (GSUEJ)

to the club namesake, John Galbraith, who was the first Dean of Engineering of the Faculty of Applied Science and

Engineering at the University of Toronto. Prof. Galbraith taught Civil Engineering and held the position of Dean of

Engineering from 1909-1914. The images drawn on the chalkboard behind the image of John Galbraith are of an E&M

field, circuit design, airplane, Rutherford’s atom model, rocketship, and a somatic cell. The enclosing atoms demonstrate

the broadening of engineering as a subject, encompassing not only Civil Engineering, but the application of physics,

math, chemistry, and biology to solve problems.

The design for the cover is courtesy of University of Toronto Libraries for the image and Sowmya Tata for the graphic

design.

About the Committee

Journal Directors:

• Zhixin Alice Ye – Engineering Science 1T4 + PEY• Areeba Zakir – Industrial Engineering 1T6 + PEY

Editors:

• Daniel Eftekhari – Engineering Science 1T6• Elana Sefton – Chemistry and Cell Biology 1T6

Reviewers:

• Esther Jang – Engineering Science 1T7• Alina Ma – Engineering Science 1T8• Matthew Ng – Engineering Science 1T6• Gengyu (Paul) Xu – Engineering Science 1T6• Tony Ye – Engineering Science 1T7• Amy Zhao – Materials Science Engineering 1T6

Professor Reviewers:

• Prof. Graeme Norval (Department of Chemical Engineering)• Prof. Hugh Liu (University of Toronto Institute for Aerospace Studies)• Prof. Adam Steinberg (University of Toronto Institute for Aerospace Studies)• Prof. Paul Yoo (Department of Electrical and Computer Engineering, and Institute of Biomaterials and Biomedical

Engineering)• Ph.D. Candidate Mukul Tewary (Institute of Biomaterials and Biomedical Engineering)

Faculty of Engineering Advisors:

• Estelle Oliva-Fisher (Assistant Director, Student Experience & Teaching Development)• Alan Chong (Senior Lecturer, Engineering Communications Program)• Susan McCahan (Vice-Provost, Innovations in Undergraduate Education, Undergraduate and Professor, Mechanical

Engineering)• Edward Sargent (Vice-Dean, Research and Professor, Electrical and Computer Engineering)

Graphic Designer:

• Sowmya Tata - Engineering Science 1T7

Layout Design:

• Ricardo Santos-Baptista - Engineering Science 1T4 + PEY

Contact Info

Galbraith Society12 Hart House CircleToronto, ON, Canada M5S 3J9Email: gsociety.uoft@gmail.comWebsite: www.galbraithsociety.ca

Directors’ Message

To our beloved readers,

It was around this time last summer when we first came up with this idea brainstorming around a table at a weekly

Galbraith Society club meeting: students were engaging in research, but getting published work was very difficult.

What if we were able to improve student communication skills as well as publishing research articles in the process?

And thus, the idea of the Galbraith Society Undergraduate Engineering Journal was born.

After months of meetings, planning, promotion, discussion, readings, and Latex formatting, we are immensely pleased to

share with you the inspiring work of young investigators conducting engineering research at the University of Toronto. A

great number of people contributed their time in helping out in this project, and we would especially like to thank Estelle

Oliva-Fisher and Alan Chong from the University of Toronto Faculty of Engineering for their support and encouragement.

In addition, we are deeply grateful to Ricardo Santos-Baptista for putting together all the articles, Sowmya Tata for her

cover design, all the authors who submitted material, as well as all of our fabulous editors, reviewers, and professors

who helped review and refine submitted articles. Lastly, we would like to thank the UTSU, Engineering Alumni Fund,

Engineering Society, and the Mechanical and Industrial Engineering Department for their sponsorship.

This year we celebrate the inaugural year of releasing the Galbraith Society Undergraduate Engineering Journal, and we

couldn’t be prouder of the result. Our journal features the work of students from topics ranging from Environmental to

Aerospace Engineering, with 6 original research articles and 3 news articles.

While reading this journal issue, we hope that you are able to immerse and intrigue yourselves into the possibilities of

research and learn a little bit more about the work of students around you.

Yours truly,

Alice Ye and Areeba Zakir

GSUEJ Journal Directors 2014-2015

Part I: Journal Papers

Evaluating Phosphate Levels in Green Roofs ofDifferent Growing MediaAllan Pretti Ogura1, Jenny Hill1, and Jennifer Drake1

For achieving sustainability, ordinary roofs have been replaced by green roofs on many buildings world-wide. Green roofs have many advantages, including the storm water retention and improved urban bio-diversity. However, green roofs have been shown to release phosphate in effluent which can increaseamount of bio-available nutrients in receiving water systems. The concentration of phosphate has beenlinked to green roof age and media type. In the city of Toronto, studies have shown that phosphate levelsare higher in the first two years of green roof operation. High phosphate levels have negative impactson aquatic environments, such as eutrophication, and thus regulating total phosphorus in green roofs isa challenge for storm water management. This study has evaluated phosphate levels in media sampledfrom 27 test beds from The Green Roof Innovation Testing Laboratory, at the University of Toronto, withdifferent combinations of growing media type, depth, vegetation and irrigation. Total phosphorus wasobtained by chemical analysis, according to Smart 3 Colorimeter procedures. Then, the obtained data wasstudied statistically, by applying ANOVA 95% of confidence, which has shown media type as the main pa-rameter for increasing runoff phosphate levels in the green roof. In this analysis, mineral media type soilextract has provided lower concentrations in comparison to organic media type. Therefore, these resultshave highlighted the importance of choosing mineral media type on a green roof for avoiding phosphatehigh levels while maintaining its benefits. This research also validated these analysis by developing aproper method of evaluating phosphate levels on green roofs.

IntroductionSustainability is the capacity of developing human ac-

tivities in order to ensure environmental conservation andto maintain natural resources for future generations [1].For this purpose, many new technologies have been de-veloped for achieving sustainability. For example, stormwater management has firstly been applied just basedon simple techniques for rainwater harvesting [2]. Ad-ditional research on sustainable buildings decided to gofurther and started to replace ordinary roofs for greenroofs on buildings of big cities for sustainable reasons.

Green roofs have many advantages, including the stormwater retention and improving urban wildlife habitat [3].Furthermore, green roofs influence micro climate, mitigat-ing urban heat island effects [4]. Green roofs also providesustainable building characteristics related to energy con-servation, since costs for cooling are reduced due to lowerambient temperatures [5]. However, green roofs havebeen shown to release phosphate in effluent [6]. Phospho-rus is a nutrient that can increase amount of bio-availablenutrients in receiving water systems, such as rivers andreservoirs. Phosphorus in natural water occurs mainly in

1Department of Civil Engineering, University of Toronto, Toronto,ON

the form of orthophosphates, condensed phosphates andorganically bound phosphates [7]. Phosphorus usually isfound as phosphate in natural ecosystems, which typicallypresent concentrations lower than 0.1 mg/L [8]. However,large amounts of phosphate are present in dischargedwastewater, because it is widely used on industrial pro-cess, fertilizers and domestic detergent [9].

High phosphorus levels have negative impacts on aquaticenvironments, such as eutrophication [10]. This phenom-ena occurs when phosphorus and nitrogen, which areessential nutrients for plants, are increasingly availablein natural streams and leading to algae blooms. Sincephosphorus is usually the limiting nutrient, water mon-itoring have to focus on phosphorus levels rather thannitrogen concentrations [11]. Therefore, regulating totalphosphorus in green roofs is a challenge for storm watermanagement, which can be assessed by proper water qual-ity analysis. In addition, green roofs should be monitoredfor preventing environmental problems while maintain-ing its benefits. Green roofs are important for storm watermanagement of sustainable cities, although they demandmore studies for making this alternative even more reli-able.

La Motte [12] has proposed two methods of evaluatingphosphate in water by using the SMART 3 Colorimeter.

2 Galbraith Society Undergraduate Engineering Journal • Volume 1, Issue 1, September 2015

Phosphate Low Range method relies on ascorbic acid re-duction with persulfate digestion. This analysis is recom-mended for drinking, surface and saline waters, but itcan also be used for domestic and industrial wastewater.The High Range method uses molybdovanate methodwith acid persulfate digestion, and it is usually appliedfor boiler, cooling, and industrial water. This study wasbased on tests beds from The Green Roof Innovation Test-ing Laboratory (GRITLab). This laboratory is located atthe University of Toronto, at the John H. Daniels Facultyof Architecture, Landscape, and Design. The GRITLab iscomposed by two main stages of study. The Phase 1 con-sists in 33 three years old test beds which have differentcombinations of growing media type, depth, vegetationand irrigation. The Phase 2 consists in organic media bedsassociated to solar panels which started operating in 2014.

Media soil sample from GRITLab Phase 2, which hasnever been irrigated before sample preparation, was eval-uated for analyzing how much dissolved phosphate isassociated to the initial flush on green roofs. Finally, thisstudy has evaluated water-extractable total phosphorouslevels in media sampled from 27 of test beds from GRIT-Lab Phase 1, which were represented in the Figure 1.Based on the literature, the concentration of phosphate hasbeen linked to green roof age and media type [13]. In thecity of Toronto, studies have shown that phosphate levelsare higher in the first two years of green roof operation[14]. Therefore, test beds from GRITLab Phase 1 are ex-pected to have phosphate concentrations in the low range,since the initial flush from the first years of operation isexpected to carry most dissolved phosphate from greenroofs. The objective of this project was to determine themajor factor that influences green roof phosphate levelsrather than replicate existing studies which have focusedon changes with age.

MethodsThe methodology was divided in four steps. First, sam-

ple collection and preparation, since soil samples mustbe diluted and extracted to produce a solution for phos-phate analysis. Second, soil media from GRITLab Phase2 was analyzed to determine maximum concentration ofphosphate from both High Range and Low Range. Third,calibration curves were created to validate the Low RangeMethod against the High Range Phosphate Analysis. Fi-nally, phosphorus total Low Range analysis applied toeach sample from the 27 studied test beds from GRITLabPhase 1.

Collecting and Preparing SamplesCollecting and preparing samples: This procedure was

prepared based on Methods of Phosphorus Analysis forSoil, Sediments, Residuals, and Waters [16], and addi-tional recommendations from Fuhrman et al. [17] on soilextraction dilution ratios and mixing times. First, soil me-dia sample from GRITLab Phase 2 and soil media samples

Fig. 1. GRITlab Phase 1 Representation (Adapted from[15].

from 27 beds of GRITLab Phase 1 were obtained with aspoon and placed in labeled tubes. These samples wereair dried overnight to remove water, so that the weight ofthe sample refers only to soil mass. In the following day,the samples were sieved to 1.70 mm, since only its smallfraction contributes to dissolved phosphate [17]. Then, foreach one of the beds, 10 g of soil was obtained and placedin a conical flask with 100 mL of Milli-Q water. Then,samples were shaken vigorously for 10 minutes. Finally,this mixture was filtered in a filter paper and the filteredsolution was used for dissolved phosphate analysis.

Phosphorus Total Low RangeThis procedure was made according to LaMotte [12]

recommendations, which is proposed based on ascorbicacid reduction with persulfate digestion method. Ideally,this experiment works in a range of 0.00 to 3.50 mg/L ofphosphate in water. First, COD reactor was preheated to150 C, the temperature that allows digestion processes. Inthe Total Phosphorus Acid Reagent Tubes, which contains2 mL of sulfuric acid 6%, 5.0 mL of sample solution and0.15 g of Digestion Reagent Powder was added in eachtube. The tubes were capped and placed in the COD for30 minutes. Then, after cooling period, 1.0 mL of TotalPhosphorus LR Hydroxide Reagent was added to eachtube. The sample at this point had no color in the blank,which was read by the SMART3 Colorimeter. By adding1.0 mL of Phosphate Acid Reagent and 0.1 g of PhosphateReducing Reagent, it was noticed color change on thesample, after 5 minutes waiting. Finally, the new readingfrom SMART3 Colorimeter represented the phosphateconcentration in the sample, reported in mg/L or ppm.

Galbraith Society Undergraduate Engineering Journal • Volume 1, Issue 1, September 2015 3

Phosphorus Total High RangeThis procedure was made according to LaMotte [12]

recommendations, which is proposed based on molyb-dovanadate method with acid persulfate digestion. Thisexperiment has a wider range, comparing to Low Range,from 0.00 to 70.0 mg/L of phosphate in water. First, CODreactor was preheated to 150C. In the Total PhosphorusAcid Reagent Tubes, which contains 2 mL of sulfuric acid6%, 5.0 mL of sample and 0.15 g of Digestion ReagentPowder was added in each tube. For this method, theblank consisted in 5.0 mL of deionized water. After allreagents were used, the tubes were placed in the COD for30 minutes. Then, after cooling period, 2.0 mL of TotalPhosphorus HR Hydroxide Reagent was added to eachtube. It was added 0.5 mL of Total Phosphorus HR Indica-tor Reagent to each one of the tubes. After a waiting timeof 7 minutes, the sample tubes were read by the SMART3Colorimeter. This value represented the phosphate con-centration in the sample, reported in mg/L or ppm.

Calibration Curves for PhosphateThis procedure was based on the guidelines from the

American Public Health Association [18], “Standard Meth-ods for the Examination of Water and Wastewater”, whichhas provided instructions for preparing phosphate stocksolution and phosphate standard solution. First, for prepar-ing a Standard Phosphate Solution, by using an analyticbalance accurate to 0.1 mg, 219.5 mg anhydrous KH2PO4was dissolved with distilled water in a 1000 mL volumet-ric flask. Then, the stock phosphate solution has presentedconcentration of 50 mg PO−3

4 /L. The desired concentra-tion was 25 mg PO−3

4 /L, which corresponded to the high-est value, and it was prepared a series of six standardswithin the required phosphate concentration range. There-fore, the recommended concentrations were 0, 5, 10, 15,20, and 25 mg/L. These six samples were tested for bothHigh Range and Low Range Phosphate Analysis. Af-ter obtained these results, the values were plotted in agraph that had a straight line passing through the origin.Comparison of both methods was proposed based on thiscalibration curves.

ResultsSoil media sample from GRITLab Phase 2 was analyzed

according to La Motte procedures for both Low Range andHigh Range. This step was crucial to determine whichthe highest concentration of dissolved phosphate in stud-ied beds, which would be expected in the initial flushof green roof operation. The results from this step weresummarized in Table 1.

Then, calibration curves were created for validatingLow Range Phosphate analysis against High Range ongreen roof beds analysis. The concentration of 25 mg/Lwas used as the highest concentration and the phosphateconcentration from a stock solution was measured accord-ing to Low and High Range Methods. The stock solution

Table 1. GRITLab Phase 2 Soil Media Sample for HighRange and Low Range Methods.

Method Phosphate(mg PO4/L)

Low Range 25.00

High Range 20.00

had initial concentration (C1)of 25 mg/L and the final vol-ume (V2) was 100 mL, so the volume of stock solution tobe added (V1) could be calculated based on the desiredfinal concentration (C2).

C1V1 = C2V2 (1)

25mg/L ∗V1 = C2 ∗ 100mL (2)

V1 = 4 ∗ C2 (3)

Table 2. Calibration Analysis.

StandardConcentration (mg/L)

Low Range(mg/L)

High Range(mg/L)

0.00 0.09 0.00

5.00 5.95 11.00

10.00 10.02 22.00

15.00 14.49 35.00

20.00 18.82 44.00

25.00 22.67 57.00

Finally, according to La Motte procedures, green roofsamples from each one of 27 studied beds from GRITLabPhase 1 were tested with Low Range Phosphate, whichhas used Ascorbic Acid Reduction Method. Furthermore,basics statistics analysis were presented in Table 3.

DiscussionFirst, based on the results presented in the Table 2, it

was possible to compare calibration curves from Highand Low Range to what was expected as a standard. Asshown in the graph from the Figure 2, Low Range Methodhas presented calibration curves relatively closer to theStandard calibration curve, providing more accuracy toits analysis and validated the initial hypothesis that LowRange would be the recommended method for evaluatingphosphate levels on three years old green roofs.

In addition, it was confirmed that age has an impor-tant hole on determining phosphate levels on green roofs.The soil media sample from GRITLab Phase 2 has pre-sented concentrations of 25 mg/L for Low Range Method.On the other hand, for three years old green roofs themaximum value was 12.42 mg/L. This can be explained

Table 3. Basic Statistics for Phosphate Analysis (mgPO4/L).

Medium Type (mg/L)

Number Average Median StandardDeviation

Compost 15 9.31 9.43 2.49

Mineral 12 4.64 3.37 3.17

Medium Depth (mg/L)

10cm 12 6.54 8.25 4.06

15cm 15 7.52 8.37 3.23

Planting (mg/L)

Meadow 15 7.10 8.20 3.54

Sedum 12 7.50 8.51 3.50

Irrigation (mg/L)

None 9 7.81 8.20 2.70

Sensor 9 6.28 8.37 3.77

Timer 10 7.32 9.35 3.78

because dissolved phosphate associates to the water thatpass through green roof beds. In very recent green roofs,the initial flush provided higher concentrations of phos-phate on the effluent. After operation time, the amountof phosphate on the effluent depends on phosphate con-centrations on storm water and on residual phosphateon green roofs. Second, by applying ANOVA with 95%confidence it was possible to create boxplots that havecompared green roof parameters and their influence onphosphate levels. These boxplots were included in theFigure 3, for depth, medium type, irrigation, and planting,respectively.

In summary, the only parameter that has shown cleardifference on phosphate levels based on ANOVA 95%confidence was media type. Mineral media type haveconsiderably lower concentrations comparing to Organicmedia type. Based on the Table 3, organic media typehad median phosphate concentration of 9.43 mg/L, whilemineral media type had median phosphate concentrationof 3.37 mg/L.

Fig. 2. Calibration Curves for Phosphate Analysis.

Fig. 3. Boxplots for Phosphate levels according to Depth,Planting, Media, and Irrigation.

ConclusionBased on the results from this study, mineral media

type has provided lower phosphate concentrations com-paring to organic media type. Therefore, for avoidingphosphate high levels on green roofs, mineral media typeconsists in a better choice. However, these phosphate lev-els are still high comparing to natural concentrations ofphosphate on natural water streams, which is the rangeof 0.01 and 0.03 mg/L [19]. Therefore, this study providesa demand on additional research for minimizing nutrientlevels on green roof effluent.

Although mineral media type has presented lower

phosphate concentrations comparing to organic media,another parameters should be evaluated for determiningthe best combination of green roof to be applied. Thischoice depends on green roof purposes. For instance, bio-diversity is more evident on organic media comparing tomineral media [20], and this is another factor that have tobe taken under consideration.

It is recommended that phosphate analysis for greenroofs should use Low Range method, since its calibrationcurve was closer to standard phosphate solution. HighRange method has presented higher sensitivity, whichcould provide lower accuracy to analysis that rely on thismethod.

Further research should evaluate nitrogen, which isalso a nutrients that can cause eutrophication. In addition,new technologies have to be developed for reducing theseenvironmental problems related to phosphate high levels.Treating this effluent or reusing the effluent water forirrigation could consist in alternatives that minimize greenroofs disadvantages.

References[1] Morelli, “Environmental sustainability: A definition

for environmental professionals,” Journal of Environ-mental Sustainability: Rochester Institute of Technology,vol. 1, pp. 225–292, 2011.

[2] UNEP, “Rainwater harvesting: a lifeline for humanwell-being,” Tech. Rep., 2009.

[3] E. Oberndorfer, J. Lundholm, B. Bass, R. R. Coffman,H. Doshi, N. Dunnett, S. Gaffin, M. Köhler, K. K. Y.Liu, and B. Rowe, “Green roofs as urban ecosystems:Ecological structures, functions, and services,” Bio-Science, vol. 57, no. 10, pp. 823–833, 2007.

[4] E. Alexandri and P. Jones, “Temperature decreases inan urban canyon due to green walls and green roofsin diverse climates,” Ed. Elsevier, vol. 43, no. 4, pp.480–493, 2008.

[5] H. F. Castleton, V. Stovin, S. B. M. Beck, and J. B. Davi-son, “Green roofs; building energy savings and thepotential for retrofit,” Ed. Elsevier, Energy and Build-ings, vol. 42, pp. 1582–1591, 2010.

[6] C. C. Glass, “Green roof water quality and quantitymonitoring,” Tech. Rep., 2007.

[7] P. C. Mishra, N. Behera, B. K. Senapati, and B. C.Guru, “Advances in ecology and environmental sci-ences.” Ashish Publishing House, pp. 215–216, 1995.

[8] W. Dodds, E. Carney, and R. Angelo, “Determiningecoregional reference conditions for nutrients, secchidepth and chlorophyll a in kansas lakes and reser-voirs,” Lake and Resevoir Management, vol. 22, no. 2,pp. 151–159, 2006.

[9] D. W. Litke, “Review of phosphorus control mea-sures in the united states and their effects on waterquality,” Tech. Rep., 1999.

[10] D. L. Correll, “The role of phosphorus in the eutroph-ication of receiving waters: A review,” Journal of En-vironmental Quality, vol. 27, pp. 261–266, 1998.

[11] D. G. F. Cunha, A. P. Ogura, and M. C. Calijuri,“Nutrient reference concentrations and trophic stateboundaries in subtropical reservoirs,” Water ScienceTechnology, vol. 65, no. 8, pp. 1461–1467, 2012.

[12] L. Motte, “Smart3 colorimeter operator’s manual,”2014.

[13] G. Harper, “Green roof water quality impacts andphysicochemical stability,” Ph.D. dissertation, 2013.

[14] TRCA, “Evaluation of an extensive green roof at theyork university, toronto,” Tech. Rep., 2006.

[15] B. Matthews and M. Perotto, Green Roofs Beds Vari-ables. Green Roof Innovation Testing Laboratory,University of Toronto, 2014.

[16] G. M. Pierzynski, “Methods of phosphorus analysisfor solid, sediments, residuals, and waters,” Tech.Rep. 396, 2000.

[17] J. K. Fuhrman, H. Zhang, J. L. Schroder, R. L. Davis,and M. E. Payton, “Water-soluble phosphorus as af-fected by soil to extractant ratios, extraction times,and electrolyte,” Communications in Soil Science andPlant Analysis, no. 36, pp. 925–935, 2005.

[18] APHA, “Standard methods for the examination ofwater and wastewater,” Tech. Rep., 1995.

[19] C. Jacobson, Water, Water Everywhere - Teacher’s Guideand Experiments to Water Quality Testing in your Class-room, 2nd ed. Loveland, CO: Hach Company, 1991.

[20] J. S. Maclvor, L. Margolis, C. L. Puncher, and B. J. C.Matthews, “Decoupling factors affecting plant diver-sity and cover on extensive green roofs,” Ed. Elsevier.Journal of Environmental Management, vol. 130, pp.297–305, 2013.

Characterization of Hydrogen Side Reaction for ZincAir Fuel CellHui Huang Hoe1 and Donald Kirk1

A zinc-air fuel cell, utilizing zinc and atmospheric oxygen as fuel, offers a cheap, environmentally-friendly yet effective alternative for energy storage and production. However, previous studies observedthe formation of hydrogen, a source of fire hazard during cell recharge, yet it was unknown whether thehydrogen evolution was significant enough to warrant design attention. To evaluate the risk quantita-tively, a sealed zinc-air fuel cell was constructed to simulate the recharge phase of a zinc-air fuel cell. Theelectric current, voltage and electrode gas evolution rate were measured. By applying material and electriccharge balance on both sides of the cell electrodes, the production rate, selectivity and composition of hy-drogen can be determined as a function of operating current and voltage. It was found that both the zincrecharge rate and the hydrogen evolution rate increased linearly with electroplating current. The result-ing hydrogen composition in the cell, above a threshold current density of 4.97 mA/cm2, increased rapidly,and then reached the asymptotical mol fraction of 2.77%, slightly lower than the hydrogen flammabilitylimit of 4%. While there is still some margin to the flammability limit, the loss of water to form hydrogenwould require the addition of water to the fuel cell, adding to the operating cost in industrial practice.From these results, a safe rechargeable zinc-air fuel cell would still be possible, paying attention to de-sign for the hydrogen side reaction issue, such as installing a hydrogen recombiner for portable-type fuelcells.

Introduction

Fig. 1. Illustration of a zinc-air fuel cell [1]

A zinc-air fuel cell is a fuel cell that draws energy fromthe oxidation of zinc metal with atmospheric oxygen toyield electrical energy. Compared with other fuel cellsystems such as lithium ion system, a zinc-air fuel cell

1Department of Chemical Engineering and Applied Chemistry, Uni-versity of Toronto, Toronto, ON

has low material cost because it does not involve preciousmetal, and its ease of construction further lowers the cost.Furthermore, the zinc-air cell does not use toxic materialssuch as lead compounds nor emit greenhouse gas, makingit very environmentally-friendly. The zinc-air fuel cell isalso versatile and has been used as alternative energy for avarying scale of operations, from the tiny button batteriesin a watch to powering automobiles and in future utility-scale power systems [2]. Such low cost, compactness, lowenvironment impact and versatility makes the zinc-airfuel cell an effective alternative for energy storage andproduction.

However, during recharge of the zinc-air fuel cell torecover the zinc metal, hydrogen gas is reported to formas by-product of reaction [5], as outlined by Table 1. Un-fortunately, the undesirable side reaction of hydrogen isthermodynamically inevitable due to its lower potential(1.229 V) than the desired zinc plating (1.59 V), as can beseen from Table 1.

The hydrogen gas formed not only wastes electricalenergy to form useless by-product, but also poses safety is-sues around the fuel cell compartment as both fire and ex-plosion hazard. Hydrogen evolution also consumes water,adding to the operating cost for the water de-ionizationsystem and the piping to constantly refill the fuel cell tankwith water.

Table 1. Competing chemical reactions during zinc-airfuel cell recharge [3] [4]

Reaction Zinc Plating Electrolysis of Wa-ter

AnodeOxygen Evolution Oxygen Evolution

4OH− → O2 +2H2O + 4e−

CathodeZinc Deposition Hydrogen Evolu-

2ZnO +2H2O + 4e− →2Zn + 4OH−

2H2O + 2e− →H2 + 2OH−

Overall2ZnO→ 2Zn + O2 2H2O→ 2H2 + O2

Eo = 1.59V Eo = 1.229V

Previous research has not evaluated the fire safety riskof such hydrogen side reaction [5], so it is unknown if thehydrogen evolution rate is actually substantial enoughto require design modification of zinc-air fuel cell. Theresearch seeks to fill the gap by quantitatively assessingthe hydrogen fire safety hazard, and to recommend actionto be taken based on the risk. The hydrogen flammabilitylimit is 4% [6], which means if the hydrogen evolution rateis high enough to have caused a hydrogen mol fractionhigher than 4% in the fuel cell compartment, the fuel cellis susceptible to fire and explosion hazards initiated byignitions such as sparks across faulty wire connection.

Methodology

Fig. 2. Diagram of the apparatus setup.

Figure 2 shows the sealed zinc-air fuel cell used to

simulate the recharge phase of zinc-air fuel cell undervarying recharge current. The cathode consists of a copperplate where both competing reactions of zinc plating andhydrogen evolution will occur. The anode consists ofa tin plate where only the oxygen evolution will occur.During electrolysis, species are depleted and hence themagnetic stirrer (Corning PC-351) is used to maintainspatially uniform electrolyte concentration throughoutthe cell.

The electrolyte was prepared by first making a 6 mol/Lsolution of potassium hydroxide solution, then add zincoxide while stirring until no more zinc oxide powderdissolves. Some zinc oxide powder was allowed to remainsuspended in the electrolyte as a buffer to maintain fixed,saturated concentration of zinc ion complex throughoutelectrolysis.

The separation affecting electrolyte resistance betweenthe cathode and the anode is measured to be 1.5cm. Theactive surface area of each rectangular electrode was mea-sured to be 31.8145cm2, which was used to normalize therecharge current to obtain current density: i = I/A.

The electrolysis was performed under constant currentcondition such that the direct current (DC) power supply(Hewlett Packard E3610A) would automatically adjustthe applied voltage to maintain the recharge current atset-point values. Constant current rather than constantvoltage operation is preferred because the recharge re-action rate is directly related by charge balance to theapplied current rather than to voltage. The mixture ofhydrogen and oxygen gas formed during electrolysis wasdirected to a bubble flowmeter where the volumetric flowrate of the mixture was measured from the time taken forthe bubble to move by certain volume.

Fig. 3. Derivation of the formulas for kinetics of both zincplating and hydrogen evolution.

Figure 3 illustrates the logic flow to derive the hydro-gen evolution rate based on the experiment measure-ments. In short, the charge balance from the rechargecurrent represents the rate of oxygen evolution while thegas flow rate measurement represents the sum of hydro-gen and oxygen evolution rate. From this, the hydrogenevolution rate could be obtained by deducting the flowrate due to oxygen gas from the total gas flow.

Finally, the mole fraction of hydrogen gas in the fuelcell compartment is calculated as:

dnH2dt

dnO2dt +

dnH2dt

This mole fraction of hydrogen represents the upper-bound of hydrogen concentration possible in the fuelcell compartment, because the compartment initially con-tained air which would dilute the hydrogen concentration.As operating time approaches infinity, the air in the fuelcell compartment would be purged out by evolved gasesand the actual hydrogen concentration would increaseand approach the maximum hydrogen concentration.

For conservative analysis, it is imperative to comparethe upper-bound hydrogen concentration to the flamma-bility limit of 4% hydrogen by volume, since the actualhydrogen concentration will always be lower than theupper-bound calculated.

ResultsThe results of the electrolysis are shown in the follow-

ing graphs:

Fig. 4. Polarization data for the zinc-air recharge system.

The important variables extracted from the graphs aresummarized in Table 2.

DiscussionFrom Figure 4, the graph of voltage against recharge

current shows a linear trend. The graph also shows thepresence of a threshold applied voltage, the electroplatingvoltage, before any current started flowing. The thresh-old voltage represents the thermodynamic barrier to beovercome before the reaction can happen.

From Figure 5 and 6, both the rate of zinc plating andof hydrogen evolution show a linear trend with respect

Fig. 5. Rate of zinc plating against recharge rate.

Fig. 6. Rate of hydrogen evolution against recharge rate.

to recharge current. The fitting of the graph also suggeststhat no hydrogen would be evolved if below a cut-offrecharge current density of 4.9 mA/cm2. The rate of hy-drogen evolution measured has a lower R2 value (0.9745)than zinc plating (0.99998) because the hydrogen evolu-tion rate was based on the gas flow rate, which has higheruncertainty than zinc plating based on current measure-ment.

From Figure 7, the resulting hydrogen composition inthe cell increased rapidly under low current and thenreached the asymptotic limit of 2.77%, slightly lower thanthe hydrogen flammability limit of 4%. The asymptotic re-sponse arises from the hydrogen evolution rate increasingslower than zinc plating rate.

limi→+∞

ai + bci + d

Fig. 7. Maximum hydrogen mole fraction attainable inthe fuel cell compartment against recharge rate.

Table 2. Important variables describing the hydrogenevolution reaction.

Parameter Value

Electroplating voltage 1.9455 V

Threshold hydrogen evolution cur-rent density

4.973 mA/cm2

Maximum hydrogen gas mole frac-tion

It can also be noted that the uncertainty increased forthe hydrogen mole fraction. This is due to uncertaintypropagation when performing arithmetic operations onboth the rates to yields mole fraction. In addition, theuncertainty of hydrogen mole fraction increases at lowcurrent density, this is because of the increased uncertaintyof using bubble flowmeter to measure extremely smallflowrates.

ConclusionThe electrolysis simulating the zinc-air fuel cell recharge

shows that hydrogen gas does not form below rechargecurrent density of 4.973 mA/cm2. While this may implythat hydrogen evolution can be prevented at low currentdensity, such low current density would require very largecell before the recharge rate is useful, making such condi-tion industrially impractical.

The maximum hydrogen concentration attainable inthe fuel cell compartment is 2.77%, lower than the flamma-bility limit of 4%. However, hydrogen evolution maybecome a problem for miniature devices where the cur-rent density has to be high for compactness and waterloss could not be replenished conveniently by piping. Inthis case, a hydrogen recombiner to form water from thehydrogen would be required. This could be done by a sim-ple catalytic recombiner or an electrocatalytic recombiner

which can accelerate the recombination rate by applyingvoltage.

AcknowledgementsThis work was supported by the University of Toronto

through the University of Toronto Excellence Award (UTEA)research scholarship. Professor Donald Kirk was respon-sible for supervising the project.

References[1] “Zinc based energy systems,” Tech. Rep., 2011.

[2] Fuel Cell Handbook. EG&G Inc., 2004.

[3] Duracell, “Zinc-air technology,” Tech. Rep., 2005.

[4] M. Carmo, D. Fritz, J. Mergel, and D. Stolten, “A com-prehensive review on pem water electrolysis,” Journalof Hydrogen Energy, vol. 38, no. 12, pp. 4901–4934, 2013.

[5] C. Lee, K. Sathiyanarayanan, S. Eom, and M. Yun,“Studies on suppression of hydrogen evolution reac-tion for zinc/air fuel cell,” Materials Science, vol. 539,no. 543, pp. 1427–1430, 2007.

[6] “Hydrogen safety,” Tech. Rep., 2015.

Investigation of Dephosphorization of FerrosiliconAlloy Subjected to Electromagnetic Levitation andArgon-Hydrogen Gas Flow MixtureAndrew Hue1, Katherine Le1, Wing Yi (Roxana) Pao1, Yindong Yang1, and AlexanderMcLean1

A current challenge facing the silicon photovoltaics (PVs) industry resides in finding economical andeffective methods for producing the 7N purity (99.99999%) pure silicon required for solar grade silicon(SoG-Si) used in PV applications. Past research has shown phosphorus to be one of the most difficultimpurity elements to remove during processing of raw silicon materials.

This research investigates the possibility of using electromagnetic levitation (EML) on commercial gradeferrosilicon samples while subjected to an argon-hydrogen (Ar-H2) gas flow mixture (48.5% H2) as a po-tential alternative to conventional processing methods used to facilitate dephosphorization of raw siliconmaterials. Inductively coupled plasma-optical emission spectroscopy (ICP-OES) was employed to detectthe changes in phosphorous concentration on the parts per million (ppm) scale. Thus far, a negative cor-relation between levitation time and phosphorous concentration has been demonstrated with a plateauin the trend occurring at 40 minutes of levitation time. A sample levitated at 20 minutes at temperaturerange of 1435 C to 1450 C demonstrated nearly a 41% decrease in phosphorus concentration; from 510ppm (unprocessed control sample) to 299 ppm. However, samples levitated under Ar-H2 gas flow for 40minutes showed a nearly negligible change in phosphorus concentration; from 299 ppm at 20 minutes to298 ppm at 40 minutes. Now that a plateau in the trend has been discovered, the next step in this researchwill be to determine whether or not EML and gas stream treatment will be effective on samples with alower initial phosphorus concentration (approximately 100 ppm).

If processing samples with a lower initial phosphorus concentrations result in a negative correlation be-tween levitation time and phosphorus impurity, these results may suggest that EML refining combinedwith an Ar-H2 gas mixture can be a promising alternative to conventional methods for producing SoG-Si.

IntroductionSolar energy is recognized as a promising, alternative

clean energy resource with the potential to satisfy theglobally distending demand for energy [1]. However,high costs, and low efficiencies, as compared to fossil fuels,have hampered progress towards widespread adaptationof photovoltaics (PVs) [2].

Silicon is the most dominant PV material in the indus-try today, being the base component of nearly 90-95% ofPVs produced [3, 4]. Silicon’s great natural abundance andproperties upon refinement make it an ideal material forsolar cells [5]. However, challenges in developing efficientsolar technologies stem from current production methodsof solar-grade silicon (SoG-Si), which are extremely costly

1Department of Materials Science and Engineering, University ofToronto, Toronto, ON

and energy intensive [4]. In order for PV to become a keyresource in the renewable energy sector, reductions in thecost of producing SoG-Si are essential [4]. Silicon amountsto over 25% of the cost to produce PV polysilicon pan-els, and approximately 40% of the cost for single-crystalsilicon systems [4].

Traditionally, silicon used in PVs are obtained fromoff-specification materials from the electronics industry,known as electronic grade silicon (e-Si). E-Si requires anultra-high purity silicon of about 99.9999999% or 9N pu-rity [4]. The relative purity of the various grades of siliconare measured in number of “nines” (N). However, thesolar industry only requires 7N (< 1 ppm impurities) pu-rity silicon [1]. Although it may appear advantageous touse off-spec e-Si, relying solely on off-spec e-Si results insupply bottlenecks [4]. Furthermore, the Siemens processused to create e-Si, is an extremely costly, toxic, energy

intensive, and slow process [4]. Hence, finding a methodto cheaply and reliably produce SoG-Si is of paramountimportance for creating PVs that are economically accessi-ble.

An alternative route that has been explored, is the useof relatively inexpensive and lower purity metallurgi-cal grade silicon (MG-Si), 98%-99% purity, as a startingmaterial to produce SoG-Si [5]. Numerous refining tech-niques have been examined to upgrade MG-Si to SoG-Si[4]. However, current methods are selective. There is nosingle method capable of removing all impurity elementsin a single step [1]. Consequently, refining MG-Si involvesa combination of methods to achieve target impurity lev-els for SoG-Si [1]. Furthermore, phosphorus and boronhave been identified as the two most difficult impuritiesto remove from MG-Si [4, 5]. It has been noted by otherresearchers that successful removal of these impurity el-ements will dictate the success or failure of producinglow-cost SoG-silicon [4].

This research focuses on studying phosphorus removalfrom ferrosilicon alloys (15%Fe- 85%Si) by utilizing a re-fining process known as electromagnetic levitation (EML)and subjecting the levitated alloy to an argon-hydrogen(Ar-H2(g) 48.5%) gas flow. The effects of time and temper-ature parameters on dephosphorization were observedand analyzed. The results of this investigation will ulti-mately contribute to the body of research involving thedevelopment of feasible processes enabling the economicproduction of SoG-Si.

MethodsMaterials and Equipment• Electromagnetic Levitation Apparatus (see Figure 2)

– Water-cooled Copper Levitation Coil – 3.2 mmouter diameter, 5-turn coil to provide electro-magnetic field needed to melt ferrosilicon sam-ples.

– Ameritherm-Ambrell High Frequency Induc-tion Heating System – to provide high frequencyAC current to copper coil; rated terminal out-put of 10 kW and frequency range 150 to 400kHz

– Aluminum Sealed Quartz Tube – to serve aslevitation chamber; 15 mm outer diameter and13 mm inner diameter

– Optical Viewing Prism – mounted on top forviewing sample and observing temperature ofmolten sample

– Two-Colour IR pyrometer (CHINO IR-CA Q3088)– to monitor the temperature of the levitateddroplet

– Copper Mould – for sample catching and dropletquenching after levitation

– Alumina Specimen Charging Rod - for load-ing ferrosilicon samples into range of copper

induction coil– Thermoplastic Tubing – for facilitating the flow

of Ar-H2 gas stream to levitation chamber

• Commercial grade ferrosilicon samples doped withphosphorus (initial concentration of 510ppm), 15%Fe-85%Si

• Hydrogen-Argon gas stream (48.5% Hydrogen)• Inductively Coupled Plasma-Optical Emission Spec-

trometry (ICP-OES) – this instrument was used todetermine trace analysis of metal elements (0.0002-1000 ppm) and a limited number of non-metallic el-ements (i.e. P) [6]. The OES instrument can measurethe relative amounts of up to 60 elements per singlesample run in less than one minute [6]. This tech-nique enabled the compositional analysis of phos-phorus in the ferrosilicon samples following levita-tion and quenching.

• Hydrofluoric (HF) Acid (with 50% HF, 50% H2O)• Pure Nitric (HNO3) Acid• Marble Mortar & Pestle for sample crushing (re-

quired for ICP-OES analysis)• Teflon beakers to hold digested samples• 15 mL and 50 mL Falcon Tubes• Electron balance

Experimental Design and ProcedureThe flow rate of the Ar-H2 gas stream was set to 500

mL/min. As shown in the schematic below (Figure 1),the argon-hydrogen gas (1) passes through the drieritecolumn (2), flowmeter (3), and flows into to the levita-tion chamber (4) where the molten ferrosilicon sampleswere levitated. The off-gases then exit via the exhaust (5)from the bottom of the levitation chamber. Figure 2 is aschematic showing the components of the EML refiningapparatus used in this research project.

Fig. 1. Gas Flow Schematic of Ar-H2 gas flow. Gas fromthe Ar-H2 tank (1) would pass through the DrieriteColumn (2) to remove any excess moisture from the gasstream. The gas stream would then flow through theFlowmeter (3) where the flow rate would be measured.The gas stream then passes through the LevitationChamber (4) to treat the levitating sample and finally exitvia the Exhaust (5).

Fig. 2. A schematic of EML apparatus. This diagramshows an already molten and levitating sample (1) underthe influence of the magnetic field from the copperInduction Coils (5). After the molten sample has beenexposed to the Ar-H2 gas stream (4) for a sufficient time,the Aluminum Platform (8) would be rotated via theRotation Knob (11) to move the Copper Mold (9)underneath the levitating molten sample. The power tothe system would then be turned-off and the coppermold would catch the falling sample. The quenchedsample could then be extracted for further analysis. [7]

After samples were levitated, they were crushed, di-gested, and prepared for analysis by ICP-OES. The marblemortar and pestle were used to pulverize levitated ferrosil-icon samples. After the levitated samples were ground toa fine powder, approximately 0.2 g of each sample weremeasured using an electronic balance and then transferredinto a Teflon beaker. Within the confines of a fume hood,3.5 mL of HNO3 acid and 3 mL of HF acid were addedinto the Teflon beakers containing the 0.2 g of samplepowder. After samples were completely digested, theywere then transferred into separate 50 mL-falcon tubesand filled with 40 mL of distilled water. The preparedsamples were then processed by ICP-OES analysis. Inaddition, a blank sample of only acid mixture (1 HF : 1HNO3) was prepared to aid calibration of the ICP-OESanalysis.

The optimal ferrosilicon sample masses to be levitatedwere determined to be within the range of 0.59 g to 0.64 g.The applied power to the levitation coil was maintainedbetween 3100 W to 4300 W. The droplet temperature rangewas 1355 C - 1720 C.

Rationale for Equipment and Technique ChoiceEML Refining

The EML process was selected for its capacity to refinemolten metals in a containerless environment [8]. Thisenvironment is advantageous because it prevents the risk

of contamination from impurities originating from a con-tainer or crucible, which may inadvertently leach into themolten sample. EML also allows for the inductive stirring,and rotation of the liquid metal through thermal gradients(convection flows), and levitational forces [8]. The EMLsystem involves passing an AC electric current through acopper coil, thus producing an oscillating current, therebycreating magnetic flux, which induces an eddy currentin the conductive material, finally causing the materialto be melted by resistive heating from the eddy currents[9]. The electromagnetic force resulting from the inter-action of the EM field, and inductive eddy current liftsthe sample [9]. Figure 3 below depicts the principle ofelectromagnetic levitation:

Fig. 3. EML Principle [10]

Ferrosilicon AlloyThe selection of the ferrosilicon alloy was based on the

activity coefficient of phosphorus in Si-Fe alloys. At acontrolled phosphorus partial pressure, Si-Fe alloys havedemonstrated enhanced dephosphorization due to theaddition of iron [11]. For specific Fe-Si compositions, in-teractions between silicon and iron result in decreasedphosphorus levels [11]. Consequently, a 15%Fe- 85%Sialloy was selected for this current research.

Argon-Hydrogen Gas StreamThe selection of argon-hydrogen gas flow was based on

steelmaking data. The interaction coefficient of hydrogenon phosphorus dissolved in an iron melt is known to beeH

P = 0.33 [12]. The positive interaction coefficient valueindicates increased activity between dissolved hydrogenand phosphorus solutes within an iron melt. Hence, ina dissolved state, hydrogen atoms interact with phos-phorus atoms and produce a repelling effect on one an-other, thereby causing phosphorus evaporation from themelt. The following thermodynamic relationships repre-sent phosphorus evaporation from iron [12]:

[P]Fe (wt%, H dissolved in Fe)⇔ 12

P2(g) (1)

Gibbs Free Energy Change, ∆G0 : ∆G0 = −RT ln K (2)

log KT = log(PP2)

aP= log

(PP2)12

fP[%P]= −8240

T+ 0.28 (3)

log fP = eHP [wt% H dissolved in Fe] (4)

eHP = 0.33 (5)

log fP = 0.33 [wt% H dissolved in Fe] (6)

Thus, it is hypothesized that hydrogen (from the Ar-H2gas stream) will dissolve in the ferrosilicon melt duringthe levitation process, interact with, and help eject phos-phorus impurities from the melt. The dissolved hydrogenis anticipated to induce a repelling effect and evaporatephosphorus from ferrosilicon.

Analysis of Time and Temperature ParametersSamples were levitated at different time intervals (6

min, 10 min, 20 min, and 40 min). This was done in or-der to map the relationship between levitation time anddephosphorization. In addition, three samples were lev-itated for 10 minutes. However, they were levitated atdifferent temperatures. This enabled analysis of the rela-tionship between levitation temperature and dephospho-rization.

ResultsThe following table outlines the results and parameters

of all successfully levitated samples.

Table 1. Results of Successfully Levitated Samples

SampleIdentifi-cation

LevitatedSampleMass (g)

LevitationTempera-

ture(C)

Amountof Time

Levi-tated(min)

AppliedPower

H1-0 Unprocessed Control Sample

H1-1 0.61 1380 10 3300

H1-2 0.64 1575 10 3100

H1-3 0.62 1450 20 3300

H1-4 0.61 1425 6 3100

H1-5 0.59 1355 10 3300

H1-6 0.59 1525 40 4281

H1-7 0.63 1700 40 3529

H1-8 0.62 1700 40 3444

Effect of TemperatureWith the melting temperature of the 15%Fe-85%Si alloy

being approximately 1600 K (1327 C) [13], levitation atincreasing temperatures of 1355 C, 1380 C, and 1575 C,was performed on three ferrosilicon samples levitated for10 minutes. It was expected that increasing temperature

would result in the evaporation and removal of phospho-rus from the molten ferrosilicon during levitation.

Fig. 4. Graph showing the relationship betweenphosphorus concentration and levitation temperature.

Effect of TimeThe effect of time on dephosphorization was examined

for eight samples levitated for 0, 6, 10, 20, and 40 minutetime intervals. An applied power range of approximately3300 W – 4300 W was used and sample mass range of0.59 g-0.64 g was employed. Note: It is expected thatthere will be a ~10% loss in silicon during levitation dueto experimental processing conditions (temperature andtime).

Fig. 5. Effect of reaction time on phosphorusconcentration in ferrosilicon.

Analysis and Discussion of ResultsEffect of Time on Phosphorus Concentrationin Ferrosilicon

Increasing levitation time resulted in a substantial de-crease in phosphorus concentration levels until approx-imately 20 minutes of levitation (see Table 1). The sam-ples were levitated in the temperature range of 1355 C- 1720 C. The unprocessed samples contained an initialphosphorus concentration of 510 ppm and the 20-minutelevitated sample contained a remaining phosphorus con-centration of 299 ppm. Hence, there was 41% decrease in

phosphorus concentration after 20 minutes of levitation.However, a limit in the trend appeared after 20 minutes oflevitation. The average concentration of three samples lev-itated for 40 minutes showed a nearly negligible decreasein phosphorus levels from 299 ppm (20-minute sample)to 298 ppm (40-minute samples). At this present study,it is unclear as to the exact cause of the limiting trend.Injecting the Ar-H2 gas stream was intended to create avacuum effect on the phosphorus partial pressure directlysurrounding the sample. In other words, the phosphoruspartial pressure in the immediate proximity of the levitat-ing sample would be extremely low due to the gas streamremoving phosphorus vapour directly surrounding thesample. This may be the first mechanism that caused thephosphorus concentration to decrease to approximately300 ppm. However, there may be a second mechanismpresent which caused the limiting trend. For example,a reaction between hydrogen and phosphorus to formhydrogen phosphide. However, this will need to be theobjective of a future study. In addition, future studiesare required to determine the efficacy of EML refiningon ferrosilicon samples with a lower initial phosphorusconcentration of 100 ppm as opposed to the 510 ppm usedin this current paper. If EML refining is still effective atlower initial phosphorus concentrations, then it is likelythat EML refining could be a potentially useful refiningtechnique for producing SoG-Si.

Effect of Temperature on Phosphorus Concen-tration in Ferrosilicon

As illustrated by the results (see Figure 4), the trendbetween dephosphorization and increasing levitation tem-perature was found to have a less substantial correlation.Based on the results of the three samples levitated for10 minutes, the ferrosilicon samples experienced a totalphosphorus concentration decrease of approximately 90ppm (from 449 ppm to 359 ppm) over the 220 C tempera-ture range (from 1355 C to 1575 C). From the resultingtrend line equation (y = −0.3861x + 966.36; Figure 4)and R2 value (0.9825; Figure 4), it is observed that theexperimental data points display a good fit. An increasein 220 C only correlated with a 20% decrease in phos-phorus concentration. Given the less substantial impactof increasing levitation temperature on dephosphoriza-tion, there is little incentive to increase temperature tofacilitate phosphorus removal. Hence, processing maybe performed at lower temperatures, which may conse-quently translate into greater processing efficiency, anddecrease expended energy. Furthermore, it is known thatthe vapour pressure of silicon increases significantly withincreasing temperature, resulting in silicon evaporationlosses at higher temperatures [1]; whereas the vapourpressure of pure phosphorus does not experience muchchange with increasing temperatures [1]. As a result, thegoal of developing a feasible silicon processing techniqueinvolves the use of lower operating temperatures, in orderto reduce the potential of evaporative silicon losses.

With the melting temperature of 15%Fe-85%Si beingapproximately 1600 K (1327 C) [13], it would be reason-able to suggest a processing temperature in the range of1300 C to 1450 C in order to account for possible super-heating that may occur.

ConclusionsDephosphorization by evaporation was observed through

the dissolution of hydrogen into molten ferrosilicon (15%Fe- 85%Si, 510 ppm phosphorus) during levitation. After 20minutes of levitation, nearly 41% of the phosphorus in theferrosilicon sample was removed; 510 ppm, unprocessedsample to 299 ppm, 20-minute levitated sample. How-ever, after 20 minutes of levitation, the decreasing trendapproaches a limit.

Increasing the processing temperature by 220 C wasfound to improve dephosphorization by only 20% (from1355 C, 449 ppm to 1575 C, 359 ppm). This suggeststhat increasing levitation temperature will have a smallerimpact on dephosphorization. Processing at lower tem-peratures is recommended as it decreases the chances ofhigh evaporation losses of silicon, and thereby, increasesthe overall process efficiency and yield.

In conclusion, future studies will be needed to deter-mine whether or not a notable decrease in phosphorusconcentrations will be observed for ferrosilicon sampleswith a lower initial phosphorus concentration (~100 ppm).If a negative trend is still detected, this would suggestthat EML refining with the Ar-H2 gas stream could be apromising alternative refining method.

AcknowledgementsWe would like to express our gratitude to both Paul

Wu and Dr. Yang who provided invaluable input andsupport during the entire research project. We would alsolike to acknowledge Professor McLean for allowing thisproject to be possible. We also wanted to extend a bigthank you to Abdol-Karim Danaei for his assistance withthe ICP-OES sample analysis. Lastly, we would like tothank the Materials Science and Engineering Departmentfor providing us with the opportunity to work on thisresearch project.

References[1] M. Johnston, L. Khajavi, M. Li, S. Sokhanvaran,

and M. Barati, “High-temperature refining ofmetallurgical-grade silicon: A review,” JOM, vol. 64,no. 8, pp. 935–945, 2012.

[2] C. Goodall, Ten Technologies to Save the Planet: EnergyOptions for a Low-Carbon Future. Vancouver: D & MPublishers Inc., 2010.

[3] A. Luque and S. Hegedus, Handbook of PhotovoltaicScience and Engineering. Chichester, West Sussex:John Wiley & Sons, Ltd, 2011.

[4] D. Lynch, “Winning the global race for solar silicon,”JOM, vol. 61, no. 11, pp. 41–48, 2009.

[5] K. Morita and T. Miki, “Thermodynamics of solar-grade-silicon refining,” Intermetallics, vol. 11, no.11–12, pp. 1111 – 1117, 2003.

[6] The University of Edinburgh, School of Chemistry,“ICP-OES and ICP-MS,” Available: http://www.chem.ed.ac.uk/research/icp.html. [Accessed: 2013Oct 29].

[7] Materials Science and Engineering, Universityof Toronto, “Electromagnetic Levitation Sys-tem,” Available: http://www.mse.utoronto.ca/Assets/MSEng+Digital+Assets/Images/Research/EmLT-schematic.JPG. [Accessed: 2013 Oct 28].

[8] D. Hectors, K. Van Reusel, and J. Driesen, “Exper-imental validation of electromagnetic-thermal cou-pled modelling of levitation melting,” Przeglad Elek-trotechniczny, vol. 84, pp. 140–143, 2008.

[9] Y. Asakuma, Y. Sakai, S. Hahn, T. Tsukada,M. Hozawa, T. Matsumoto, H. Fujii, K. Nogi, andN. Imaishi, “Equilibrium shape of a molten silicondrop in an electromagnetic levitator in microgravityenvironment,” Metallurgical and Materials TransactionsB, vol. 31, no. 2, pp. 327–329, 2000.

[10] M. Popa, “Study of an electromagnetic levitation sys-tem,” Nonconventional Technologies Review, no. 1, pp.34–38, 2010.

[11] S. Ueda, K. Morita, and N. Sano, “Thermodynamicsof phosphorus in molten Si-Fe and Si-Mn alloys,”Metallurgical and Materials Transactions B, vol. 28,no. 6, pp. 1151–1155, 1997.

[12] M. Hino and K. Ito, Thermodynamic Data for Steelmak-ing. Sendai: Tohoku University Press, 2010.

[13] École Polytechnique de Montreal and McGillUniversity-Facility for the Analysis of Chem-ical Thermodynamics (FACT), “Fe-Si- FACT-Sage light metal alloy databases,” Available:http://www.crct.polymtl.ca/fact/phase_diagram.php?file=Fe-Si.jpg\&dir=FSlite. [Accessed: 2014 Jan21].

An Evaluation of the Stability of the Boeing 747Flight Simulator Model using a CNBdynamic

Arthur Brown1 and Peter R. Grant2

Aircraft are designed to be stable throughout normal flight operations; for example, during straight-and-level flight, climbs and descents, and turns. However, a stalled aircraft often becomes laterally unstable;that is, any small lateral disturbance will build over time, becoming either a dangerous spin or spiral dive.Flight simulators can be useful in training pilots for these situations; however, this requires the flight sim-ulator model to accurately reflect post-stall aircraft behavior. The aim of this research was to evaluate theUTIAS flight simulator model for the Boeing 747 to determine under what conditions it is laterally un-stable. This was done by studying an aerodynamic term called CNβdynamic

, which measures lateral static

stability. In general, if the aircraft is laterally statically stable, CNβdynamicis positive; if not, it is negative.

The fully nonlinear MATLAB/Simulink model was executed multiple times, using different flight condi-tions. Data from the iterations was used to create a 2D map on which CNβdynamic

was evaluated and graphed;

regions of the map where CNβdynamicwas negative indicated lateral instability. The experiment showed that

the model becomes less laterally stable as α (angle of attack) increased, with CNβdynamicbecoming negative

at around α = 22 C. Angle of attack is the angle in the vertical plane between the relative wind and theaircraft’s longitudinal axis. CNβdynamic

became positive again at around 30 , and remained positive up to

60 . The 747 model is therefore statically unstable for 22 < α < 30 . One limitation of this research isthat the simulator model is fully nonlinear, while CNβdynamic

assumes a linearized flight model. This can

lead to modeling errors. However, CNβdynamicis known to be a good predictor of lateral departure even at

high angles of attack.

IntroductionMost aircraft are designed to be stable throughout the

course of normal flight operations, but they can becomeunstable under other conditions. For example, many air-craft become laterally unstable at higher angles of attack,at or near stall (see Figure 1 for a diagram showing angleof attack). It would therefore be beneficial to train pilotsto handle the loss of lateral stability associated with stallusing flight simulators. This requires that the simulator’sflight model accurately reflects stall and post-stall aircraftbehavior. The purpose of this research was to evaluatethe lateral static stability of the UTIAS flight simulatormodel for the Boeing 747, to see how effectively it modelsstall conditions. This was done by studying a parametercalled CNβdynamic

. CNβdynamicasks the question: if the aircraft

is laterally perturbed from a reference condition of steady,symmetric flight, are the resulting forces and moments

1Division of Engineering Science, University of Toronto, Toronto, ON2Institute for Aerospace Studies, University of Toronto, Toronto, ON

in the correct direction to oppose the perturbation (i.e. tobring the aircraft back to its reference condition) ? If this isthe case, then the aircraft is laterally statically stable andCNβdynamic

is positive; if not, it is negative.

Aircraft flight is normally described by two sets of fourcoupled ordinary differential equations, linearized abouta reference condition (typically where the aircraft is insteady, level flight). One set describes lateral behavior,and the other set describes longitudinal behavior [1]. Bothsets of equations can be given in state-space form:

x = Ax + Bc (1)

X is the state vector, c is the control vector, and A and B aresystem matrices. Note that since the system is linearizedabout a reference condition, A and B are constant matrices.The full details of this system are beyond the scope of thispaper, but a full derivation and discussion is presented in[1]. Since only lateral stability is of interest here, only thelateral equations are required. Also, static stability of anaircraft must be evaluated in the absence of control inputs

[1]; c is therefore 0. We are left with the system:

x = Ax (2)

Fig. 1. a diagram showing α (angle of attack) and β(angle of sideslip). Both are angles between the relativewind (the direction, relative to the air, in which theaircraft is moving) and the aircraft’s longitudinal (x) axis.This diagram was originally Figure 1 from [2].

This is a 4x4 system of first-order linear ordinary dif-ferential equations with constant coefficients, which canreadily be solved using eigenvalue and eigenvector analy-sis. The solution to equation (2) is of the form:

x(t) = k1u1eλ1t + ... + k4u4eλ4t (3)

ki are constants determined by the initial conditions, λiare the eigenvalues of A (real and/or complex), and uiare eigenvectors. In order for the airplane to be staticallystable, all of the eigenvalues must have negative real parts.If this is the case, it can be seen from Equation (3) thatx(t) will converge to a constant; if not, x(t) will diverge.By applying Routh’s stability criteria to the state-spacesystem of Equation (2), it can be shown that a necessary(but not sufficient) condition for all of the eigenvalues ofA to have negative real parts is for CNβdynamic

to be positive

[3]. For this reason, CNβdynamicis also known as the lateral

divergence parameter. A full derivation and discussionof CNβdynamic

is included in [3], and the equations used to

calculate CNβdynamicare given below:

CNβdynamic= [CNβ

]′cos(α0)− [CLβ

]′sin(α0) (4)

[CNβ]′=

I = Ix Iz − [Izx]2 (7)

α0 is the steady-state value of α (angle of attack); see Fig-ure 1 for a diagram. Ix and Iz are moments of inertia aboutthe X and Z axes respectively, and Izx is a product of iner-tia. See Figure 2 for a diagram of the axes. CNβ

and CLβare

called aerodynamic derivatives in coefficient (nondimen-sional) form; they represent approximately how L (rollingmoment) and N (yawing moment) change with β (angleof sideslip). In order to linearize the flight model, it mustbe assumed that CNβ

and CLβare constants. See Figure

1 for a diagram showing β, and Figure 2 for a diagramshowing L and N.

[ 12 ρV2Sb]

; CLβ=

[ 12 ρV2Sb]

Nβ∼=

∂N∂β

; Lβ∼=

∂L∂β

ρ is the air pressure, V is the aircraft speed, S is the surfacearea of the wing, and b is the wingspan. The nondimen-sional coefficients are used for mathematical simplicity;dividing through by 1

2 ρV2Sb means that they have nounits, so they are unaffected by changes in ρ or V.

Fig. 2. A diagram of the standard system of axes (bodyaxes) used to model aircraft stability and control; rollingmoment L and yawing moment N are also shown. Thisfigure was originally Figure 1.6 from [1].

Although the simulator model is fully nonlinear, itis necessary to linearize the system in order to deriveCNβdynamic

. This can result in errors at higher angles of at-

tack, where the linear assumption is less valid. However,CNβdynamic

is known to be a good predictor of lateral de-

parture even at high angles of attack [2], [3]; part of thepurpose of this research is to determine whether CNβdynamic

gives reasonable predictions for the 747 model. Note thatalthough CNβdynamic

> 0 alone does not guarantee lateral

static stability, it is known to give good predictions formost aircraft of conventional configuration; as such, it hasbecome a standard tool in the aircraft design process [2],[3]. This is elaborated further in the Discussion section.The first step in obtaining CNβdynamic

is to trim the model

to fly straight and level; this condition then serves as thereference condition about which equation (2) is linearized.CNβdynamic

can then be calculated by laterally perturbing

the model (in β), recording the resulting changes in rollingand yawing moment, then applying equations (4) through(9). A more detailed description of the procedure is pro-vided in the next section.

MethodsThe procedure for the experiment was as follows:

1. A duplicate Aerodynamics block was inserted inthe MATLAB/Simulink flight simulator model, insuch a manner that its angle of sideslip could be setexternally.

2. The model was set to trim to a certain angle of attackαtrim; it trimmed for 5 seconds. In this context, trim-ming means setting the model’s flight controls suchthat it is in steady level flight at the given αtrim. Thiswas done automatically using the model’s trimmingblock.

3. After the model was finished trimming, the dupli-cate Aerodynamics block was laterally perturbed byexternally modifying the sideslip angle βperturb.

4. Values of L and N right after the perturbation wereobtained from both the original and duplicate block.The difference was taken; dividing through by βperturbyielded Lβ and Nβ. Equation (8) was then used tocalculate CLβ

and CNβ.

5. CNβdynamicwas calculated using Equations (4) through

6. Steps 2 through 5 were repeated while varying αtrimand βperturb.

CNβdynamicwas calculated for −10 < αtrim < 60 and for

1 < βperturb < 20

ResultsTwo plots were generated:

• A 2D map (Figure 3) showing with dots where CNβdynamic

is positive or negative, with αtrim on the x-axis andβperturb on the y-axis.

• A 3D surface plot of the same (Figure 4), which givesan impression of the magnitude of CNβdynamic

as op-

posed to a simple positive/negative graph.

Fig. 3. the CNβdynamicmap. Blue dots indicate positive

CNβdynamic; red dots indicate where it is negative.

Fig. 4. The CNβdynamicsurface plot. Green regions mark

positive CNβdynamic; red regions show where it is negative.

The yellow marks the boundary (i.e. where CNβdynamic∼=

The map and surface plot showed thatCNβdynamicwas pos-

itive for α < 20 . It became negative for approximately22 < α < 30 , then became positive again up to 60 . The

CNβdynamic> 0 criterion therefore predicts lateral static in-

stability in the region 22 < α < 30 , and static stabilityelsewhere. This trend was more or less consistent acrossthe different values of βperturb.

DiscussionAs it is already known that aircraft tend to be laterally

unstable at stall, the results indicated that the stall angleof the 747 is about α = 22 The 747 model is therefore onlylaterally unstable at or near stall. It makes sense for the747 to be laterally unstable at stall, for two reasons:

1. Close to stall, the vertical stabilizer is masked by thewing wake, destabilizing the aircraft.

2. Wing sweep effects (which stabilize the aircraft) arediminished near stall.

As stated in the Introduction, it is important to rememberthat CNβdynamic

> 0 alone does not guarantee lateral static

stability. However, CNβdynamicis known to give good predic-

tions for most aircraft configurations. This was discussedin a 1972 NASA paper entitled “Summary of DirectionalDivergence Characteristics of Several High-PerformanceAircraft Configurations [2],” in which CNβdynamic

data ob-

tained from both force tests and free-flight model testswas correlated with lateral directional divergence of wind-tunnel models. The correlations were rated as good, fair,and poor. The rating of good was assigned to configu-rations for which CNβdynamic

correctly indicated no diver-

gence or correctly indicated a divergence and the angleof attack at which it occurred. The fair and poor ratingswere used respectively to designate configurations forwhich CNβdynamic

either fails to predict accurately the angle

of attack at which a divergence occurs or fails to predict adivergence altogether. The correlation was good in nearlytwo-thirds of the cases. This shows that CNβdynamic

can be

a useful tool in predicting the occurrence of directionaldivergence [2].

The pattern of CNβdynamicbecoming negative at around α

= 20 then becoming positive again at a higher α has beenobserved for other aircraft with similar configurations.Figure 11 of [2] shows a model of a swept-wing aircraftin which CNβdynamic

was calculated as positive, except for

19 < α < 23 . The results obtained here for the 747 aresimilar. The paper also notes that “CNβdynamic

correlates

fairly well with [lateral] divergence,” indicating that thelinearized flight model is still useful for predicting lateralstatic stability.

One limitation of this research is that the model hadinsufficient thrust from the engines to achieve straightand level flight for angles of attack above about 40 . Theaircraft was therefore descending when it was perturbed

to calculate CNβdynamicat these high angles of attack. How-

ever, according to [3]: “gravity should not directly cause alateral-directional aerodynamic instability.” Therefore, thesign of CNβdynamic

(and hence the measurement of lateral

static stability) should be unaffected.

ConclusionThe lateral static stability of the Boeing 747 flight sim-

ulator model has been evaluated using a CNβdynamicmap.

It was discovered that the model becomes less laterallystable as αtrim increases; CNβdynamic

becomes negative at

around αtrim = 22 . However, CNβdynamicbecomes positive

again at around αtrim = 30 , and remains positive up to60 . The 747 model is therefore statically unstable for 22

< α < 30 . The research also shows that CNβdynamicis an

effective predictor of lateral static stability for aircraft ofconventional configuration, despite its dependence on alinearized flight model. The methodology used here cantherefore be applied to nonlinear flight simulator mod-els of other aircraft with similar configurations, to deter-mine whether they accurately reflect post-stall behavior.It would be interesting to apply the methodology to air-craft of other configurations. For example, what doesCNβdynamic

tell us about aircraft with T-tails? Aircraft with

T-tails have the horizontal tail (as well as the vertical tail)masked at stall; this suggests that they might not becomestable again at higher angles of attack.

AcknowledgmentsThe author would like to thank Professor P. R. Grant,

his supervising professor and head of the Vehicle Simu-lation Lab at the U of T Institute of Aerospace Studies.The author would also like to acknowledge the adviceof Rakibur Rahman, his colleague and researcher in theVehicle Simulation Lab.

References[1] B. Etkin and L. D. Reid, Dynamics of Flight: Stability

and Control, 3rd ed. Danvers, MA: John Wiley & Sons,1996.

[2] H. D. Greer, “Summary of directional divergence char-acteristics of several high-performance aircraft config-urations,” Tech. Rep., 1972.

[3] F. H. Lutze, W. C. Durham, and W. H. Mason, “Unifieddevelopment of lateral-directional departure criteria,”Journal of Guidance, Control, and Dynamics, vol. 19, no. 2,pp. 489–493, 1995.

The Design and Testing of a Student-BuiltParrafin-Aluminium-Nitrous Oxide Hybrid SoundingRocketJeffrey R. Osborne1,2, Ashis Ghosh1,3, Adam De Biasi1,3, Hayden Lau1,4 and JeremyChan-Hao Wang1,4

This paper presents the final design, testing methods, and results for the University of Toronto AerospaceTeam’s (UTAT) sounding rocket, the Eos II. The rocket was designed over a period of 10 months and wasdesigned to reach an apogee of 3km while carrying a 4.5kg payload. Eos II was powered by a 13 000-Nshybrid rocket engine that used a mixture of paraffin wax and aluminum as fuel with nitrous oxide asthe oxidizer. Many of the overall system components, including engine, aerodynamics, structural mem-bers, and flight performance were simulated through both in-house and commercial software packages.Ground tests were performed to verify these predictions.

IntroductionIn recent years, the aerospace industry has demon-

strated growing interest in hybrid rocket technology dueto advantages in safety and cost. In hybrid rockets, the ox-idizer and fuel are in different states of matter—typicallyliquid and solid, respectively. Hybrid propellants are keptisolated until the engine reaction begins, and the reactioncan be throttled by varying liquid oxidizer flow. By con-trast, solid motors have premixed solid propellants thatmay ignite without warning, and which will continue re-acting until fuel or oxidizer has been depleted. Liquidengines on the other hand retain propellant isolation butcome with complex and heavy plumbing [1] [2]. Currently,organisations like NASA and the Space Propulsion Groupare developing hybrid technologies to enable future atmo-spheric research and high-altitude deployments. Whilehybrid rockets retain the mechanical simplicity of solidsand the safety of liquids, these advantages come at theexpense of extreme operational complexity [3].

At the University of Toronto, a team of undergradu-ate and graduate students with the University of TorontoAerospace Team (UTAT) Rocketry Division applied a com-plete systems approach to the development of hybridsounding rocket technology, with Eos II being the mostrecent output of UTAT Rocketry’s activities. The primaryobjective of this project was to implement a hybrid rocketlaunch platform for deploying 4.5kg payloads to 10 000ft.Unlike other hybrid rockets, the design of Eos II empha-

1University of Toronto Aerospace Team, Toronto, ON2Institute for Aerospace Studies, University of Toronto, Toronto, ON3Department of Mechanical and Industrial Engineering, University

of Toronto, Toronto, ON4Division of Engineering Science, University of Toronto, Toronto, ON

sized ease of manufacture and assembly, high modularity,and low-cost but effective solutions. This paper focuses onthe propulsion and fluids subsystems, with supplemen-tary sections on the related structures and avionics sys-tems. The design approach taken emphasized the use ofsimulation tools—both in-house and commercially avail-able—which were compared with experimental testing.

MethodsThe Eos II rocket used a custom 13,000Ns hybrid rocket

engine dubbed the “Bia II”. The fuel was a mixture ofparaffin and aluminum powder, and the oxidizer wasnitrous oxide (N2O). Key features of the Bia II chamberwere pre- and post-combustor, wagon-wheel fuel coregeometry, CFD-modelled injector assembly, and nozzlecooling system.

MATLAB Simulation ToolBia II was designed with an in-house MATLAB pro-

gram to simulate engine performance given key designparameters. The most promising handwritten designswere inputted into the code and then iterated on to refinefinal predicted performance. This tool was used in thedesign of the injector plate, fuel core geometry, and nozzlegeometry. Important elements of this program include achemistry model that predicts combustion chamber tem-peratures, a thermo-physical model of nitrous oxide emp-tying from the flight tank, 2D regression-rate predictionsfor wagon wheel fuel core geometry, nozzle loss factorsaccounting for non-uniformities in the fuel core geometry,feeding system loss factors, and increase in throat diame-ter due to gradual nozzle erosion. The basic flow for the

calculations is adapted from known calculation methods[4] and shown in Figure 1, along with the final predictedperformance of the rocket.

Fig. 1. Calculation flowchart and predicted performanceof Eos II, from the in-house MATLAB Simulation Tool.

Injector Plate and Pre-combustorThe injector plate and pre-combustor dispersed oxi-

dizer into the injector. Three main design criteria wereof interest to the injector: (1) high discharge coefficient tomaximize oxidizer flow rate; (2) pressure drop must be atleast 20% of combustion chamber pressure to preventbackflow during combustion pressure instabilities; (3)high flow evenness to support oxidizer atomization andmaximize surface area for reaction. To analyse these com-peting effects, a Computational Fluid Dynamics (CFD)study was undertaken in ANSYS. Since droplet size couldnot be effectively modelled, flow evenness was taken asthe parameter to assess atomisation [4]. Three base de-signs representing three distinct injector designs werechosen for the study (Figure 2). Instead of using nitrous,which poses various thermophysical challenges in mod-elling [5], CFD was performed assuming incompressibleair as the working fluid with no-slip condition at the walls;a comparison was then made to cold-flow testing. It wasunderstood that the use of this would not provide reliablequantitative values for injector performance, but would

allow for a useful qualitative comparison between injectorgeometries. By varying mesh resolution, the CFD resultswere convergent and independent. The simulations per-formed were 3D, using an incompressible steady-statesolver with a k-ε turbulence model. 2D velocity contoursfor the three cases are shown in Figure 2. Note that theevenness coefficient Ce is defined as the following, whereσU is the standard deviation of the fluid speed and y f c isthe y-position at the top of the fuel core:

Ce =1σU|y f c (1)

The chamfered holes injector (Case II) was selectedbecause of its relatively higher discharge coefficient anddownstream flow evenness.

Fig. 2. Three injector geometries compared in the CFDstudy, along with the CFD velocity contours. The primarycomparison metrics were pressure drop (must be at least20% of combustion chamber pressures to preventflashback), discharge coefficient, and downstream flowuniformity.

Combustion Chamber and Fuel CoreThe fuel core geometry was selected as a wagon wheel

geometry to increase thrust at take-off when the launchvehicle is most massive. An optimization algorithm builtinto the in-house simulation was used to select a shapewhich would provide maximum altitude, with the finalgeometry shown in Figure 3. In this algorithm, the cross-sectional length of the wagon wheel “spokes”, which af-fected average exposed fuel surface area, was iteratedupon until maximal engine pressure was obtained. Manu-facture of the fuel core was achieved by means of a foampiece that was cut with a CNC hot-wire to bear the wagonwheel cross-section, which was placed in the engine. Af-ter fuel was poured in, the core was then dissolved withacetone to leave the final profile seen once again in Figure3.

Post-combustorAfter initial engine testing (explored further in Engine

Testing Summary under Key Results and Discussion),

Fig. 3. (Top Left) Final fuel core wagon wheel geometrydesign, (Top Right) fuel core foam mandrel, and (Bottom)final fuel core with igniters already in place.

traces of fuel discovered on the nozzle suggested thatchamber residency time was insufficiently high to allowreactants to combust inside the engine. Consequently,a ‘post-combustor’ was installed—this was effectively avoid space downstream of the fuel core, and provided ad-ditional time for propellants to mix and react. To protectthe aluminum walls from high heat fluxes in this region,an ablative composite liner was used which consisted offiberglass held together with a urea-formaldehyde glueresin.

Nozzle Assembly and Cooling JacketThe Bia II nozzle was a deLaval nozzle with conical

divergent section, constructed from graphite due to itsfavourably high temperature resistance. As a result ofgraphite’s brittle properties—which could cause the noz-zle to fracture during uneven loading from engine start-upand shut-down transients—a stainless steel ‘backing-plate’support was made to hold the nozzle securely in place.Furthermore, to protect this plate against melting inducedby graphite’s high thermal conductivity and the intenseheat fluxes in this region, a cooling jacket was designedwhereby the oxidizer was passed through the nozzle as-sembly prior to entering the combustion chamber, coolingthis assembly (see Figure 4 for the final nozzle assemblyand cooling jacket). A 2D steady-state CFD simulationwas performed on the cooling jacket in order to deter-mine its performance, with the resulting contours shownin Figure 4. The simulation predicated that the coolingjacket would chill the backing plate to 1300C, a temper-ature at which steel would be softer but not yet havemelted. Additionally, the predicted temperature rise ofnitrous oxide across the cooling jacket was small due tothe extremely brief time that fluid would remain insidethe jacket. Pressure drop across the jacket was mitigatedthrough iteration, which brought it down to a predicted40psi at engine start-up when pressure losses would begreatest.

Oxidizer SubsystemThe role of the oxidiser subsystem was to house the

oxidiser supply while the rocket rested on the launch pad,and to purge the oxidizer tank when the engine was ig-nited. The oxidizer system consisted of 3 main parts: the

Fig. 4. (Centre) Nozzle and cooling assembly showinggraphite nozzle, cooling jacket, and support structure.Note that the cooling jacket was designed to cool not thenozzle, but the stainless support structure; hence thecooled region was not located near the nozzle throatwhere the higher heat fluxes would occur. (Left) Velocityand (right) temperature contours of flow through nozzlecooling jacket. Flow runs from left to right in theseimages. Velocity ranges from 0m/s (blue) to 30m/s (red);temperature ranges from 20.0 (blue) to 20.4 (red).

oxidizer tank, the plumbing assembly, and the oxidizer ac-tuation valve. The first iteration of the oxidizer tank failedunder hydrotesting due to the use of unevenly loadedgraphite gaskets that held in the tank caps. A second iter-ation replaced these graphite gaskets with welded caps,which also failed at around 500 psi. Diagrams of thesetanks are shown in Figure 5. Due to time constraints, athird iteration was never made; instead, an aluminumindustrial cylinder that was used for ground testing wasused in flight. This tank had a slightly lower capacity thandesired (5000cc versus 5500cc), and was slightly heavierthan the designed tank. This tank was rated to 2000psiand numerous hydrotests demonstrated its operationalsafety. In addition, the oxidizer feeding system housedvarious items to improve safety and reliability of the ox-idizer system, with the final rendering shown in Figure5. The feeding system included a pressure relief valve setto 1000psi, a manual relief valve, a pressure transducer, athermocouple, and a quick connect for tank filling. Finally,the actuation valve enabled the oxidizer feeding systemto be switched between three states: vent, off, and main.This would allow the vent line or the main line to flowthrough the valve, as well as turn off the valve entirely.These requirements were met by means of a custom elec-tromechanical system that actuated a commercial 3-wayvalve, with the final rendering shown in Figure 5. Thevalve was actuated with a DC motor with a built-in en-coder (The motor used is a DC brushed motor with a built-in 131:1 gearbox and 64 CPR hall-effect-based encoder).The torque required to open the valve was substantial,

Fig. 5. (Top left) Final rendering of the first iterationflight tank, showing bolt circle and vent tube. (Top right)Deformed flanges as a result of over-tightening of boltring in order to accommodate poor seal provided bygraphite gaskets. (Bottom left) Final rendering of thesecond iteration flight tank, showing vent and main lines.(Bottom right) Failed weld, which occurred at 500psiduring hydrotesting.

requiring a large motor. A DC motor with a stall torqueof 250oz-in was selected, which provided a valve open-ing safety factor of 2.4 against measured torque valuesrequired to open the valve.

Key Results and DiscussionInjector Plate Cold Flow Testing

In order to verify the injector simulations, a cold flowtesting campaign with incompressible air as the workingfluid was undertaken on Case I in order to verify its re-sults. This was achieved by placing connecting the injectorassembly to a 100 psi compressed air line, and assumingthat exit pressure was approximately atmospheric. Massflow rate was measured by way of a Coriolis mass flowmeter. Doing so, Case I had a discharge coefficient of 0.26,giving an error of 7%. Following this it was assumed thatthe CFD simulations were yielding adequately accurateresults to limit the need for further cold flow testing. Onefinal iteration on the injector was made following enginetesting (see Engine Testing Summary), when it was discov-ered that the residency time of the combustion chamberwas too low to permit complete combustion. The finalinjector included chamfered holes on the downstreamside to reduce injection velocities and thereby increaseresidency time. No CFD study was undertaken on thisinjector due to time constraints. A cross-section of thefinal injector is shown in Figure 7.

Fig. 6. (Far left) Final rendering and (left) actual design ofthe oxidizer feeding system. To improve safety andredundancy, the feeding system included a pressure reliefvalve, manual relief valve, pressure transducer,thermocouple, and quick connect feeding line. (Right)Final rendering of the oxidizer actuation assembly and(far right) final assembly, as seen in oxidizer feedingsystem.

Engine Testing SummaryThe Bia II engine underwent 4 static test fires, with the

results summarized in Table 1. This was accomplished byinverting the engine and bolting it into a concrete blockoutdoors, so that thrust pointed downward and the en-gine remained fixed in position. Oxidizer tank and as-sociated flow components were hauled outdoors—in theinterest of conserving a limited nitrous reserve, all testswere conducted with lower total volumes of oxidizer thanwould actually be used during flight. Although compre-hensive testing on the Bia II engine was not completed,the testing revealed a number of key results.

Fig. 7. Final injector design had similar geometry as inCase II, but included downstream chamfered holes toreduce injector velocity and thereby increase combustionchamber residency time. No CFD study was undertakenon this injector due to time constraints.

Postcombustor and Cooling JacketThrough informal sub-scale testing using a 2” inner di-

ameter steel pipe containing paraffin and postcombustorlining, an experimental average erosion rate was extractedin the range of 0.01-0.04in/sec. This is comparable to othermodern ablative liners with erosion rates in the range of0.0005 (carbon-fibre) to 0.02 (silica cloth phenolic) [6]. Theaddition of the post-combustor to the Bia II increased en-gine performance by 25% (see Table 1) with a 5% increasein vehicle mass, thus justifying its use. Moreover, the cool-ing jacket held engine performance relatively constant(The slight increase in performance of Test 4 as comparedwith Test 2 is likely a combined effect of an increase in

performance from a longer fuel grain, and a decrease inperformance as a result of the increased pressure dropacross the feeding system on account of the cooling jacket),and although no temperature measurements were taken,no system failure occurred.

Engine Performance Data

Fig. 8. Plot of oxidizer tank pressure, combustionchamber pressure, and engine thrust during Test 4. Anengine burn can be separated into 3 distinct phases;ramp-up, steady-state, and tail-off.

The engine performance data itself can be further ex-amined. Using Test 4 (Figure 8) as a case study, the engineburn may be sectioned into 3 distinct regions: ramp-up,steady-state, and tail-off. For the ramp-up period, the rateof increase of chamber pressure is of particular interest.This is because the slope of the pressure curve during theramp-up period is dependent upon the turbulent flamepropagation speed ST down the engine, as given by:

ST =` f c

t2 − t1(2)

where t1 and t2 are respectively the time of the startof the ramp-up period and the start of the steady-stateperiod. In this case, it was assumed that the ignition ofthe three B-class solid motors was sufficient to vaporizeenough paraffin to allow such flame propagation. Usingthis, flame speed may be calculated to be 50 cm/s. The-oretical laminar flame speed values for paraffin are onthe order of 30-40 cm/s [5]. A number of reasons couldaccount for this difference. The main reason is likely thepresence of turbulent flow inside of the engine, whichwould act to increase the flame propagation speed. Thepresence of aluminum powder may have affected flamespeed as well, but to an unknown degree. To gain arough idea of the turbulent levels inside of the engine,Damkohler’s relation for turbulent flame speeds may beapplied [7]:

= 1 +(v′rms)

which after rearranging and the inclusion of the bulkfluid velocity U (determined from simulations where ithas been assumed that the flow is in the vapour state oncereaching the engine), it can be said that:

v′rmsU≈ 0.03 (4)

indicating that inside the engine there is low turbu-lence intensity. One thing that can be concluded fromthis is that the ramp-up period can likely be reduced byincreasing the turbulence generated at the injector. Asfor steady-state operation, it is evident that although thechamber pressure rises during steady-state operation, thethrust during this period does not change. The increase inchamber pressure can be ascribed to only to an increasein the fuel mass flow rate due to increasing burn area, asthe oxidizer mass flow rate during this period is decreas-ing indicated by decreasing oxidizer tank pressure. Thiscauses O/F ratio to decrease, which reduces chamber tem-peratures and thus exhaust velocity (Equation 5 below).However, the increased engine pressure raises exit planepressures (Equation 6).

√TRMγ

1 + (γ−1)2 M2

Pt= (1 +

γ− 12

M2e )− γ

(γ−1) (6)

Therefore the changing propellant mass flow rates, de-creasing exit velocity, but increasing exit plane pressurecould conceivably result in non-changing thrust, as seenin the thrust equation:

F = mve + (Pt − Pe)Ae (7)

However, it is strongly suspected that the MATLAB sim-ulation contained an erroneous thermochemistry model,causing values for predicted chamber temperature, prod-ucts species concentrations, and specific heat ratio valuesto be incorrect (compare Table 1 with Figure 1). The criti-cal assumption of this approach was that the compositionof the products was fixed, which is likely invalid. Usingthis method predicted chamber temperatures on the orderof 2750, and product molar mass on the order of 40g/mol.Compared to NASA’s Chemical Equilibrium with Appli-cations, which predicts chamber temperatures and prod-uct molar masses on the order of 1800 and 15g/mol re-spectively [7], a clear discrepancy is evident. Moreover,NASA CEA predicts product species not represented inthe above chemical equation, including large fractionsof CO, H2, AlN, C, and HCN. The tail-off period is de-noted by when the liquid nitrous supply in completelyconsumed, and the thrust quickly drops as vapour nitrousis now consumed. Consistent in every test performed onthe Bia II engine is the characteristic spike in performancejust at the onset of the tail-off period. It is strongly sus-pected that the vent line in the flight tank retained some

Table 1. Tabulation of Bia II engine static testing performance

Test Nitrous Level Peak Sustained F [lbf] Peak Sustained Pchamber [psi] PC? [Y/N] NC? [Y/N]

1 0.50 101 102 N N

2 0.25 125 136 Y N

3 0.25 N/A N/A Y N

4 0.25 135 142 Y Y

quantity of liquid oxidiser from pre-testing liquid oxidiserfilling procedures, which was released at the end of theengine burn to produce the momentary spike in thrust.

Future RecommendationsThe MATLAB chemical model could be entirely re-

placed by a sufficiently accurate multivariable polynomialregression across inputs and outputs of NASA CEA. Fu-ture injector CFD could employ a more pressure-gradienttolerant model, instead of the more standard k-ε usedhere. The Injector design could be modified such that theinjection velocity matches the flame propagation speed.The injector can also be designed to increase turbulencelevels, such as by a swirl injector, given that current tur-bulence levels inside the engine appear to be low and notconducive to mixing. The current conical nozzle couldbe replaced by a more efficient bell nozzle, which can bedesigned with readily available guidelines [5]. Finally,the ability to throttle the engine could be implementedby varying the position of the actuation valve in orderto control the flow of nitrous, which would aid in reach-ing the exact target altitude necessary for IREC. However,extreme care would have to be taken in doing this; it isunknown how the engine will perform at various throttlelevels. It is possible that at a specific throttle setting thecombustion would become thermoacoustically unstableand lead to a catastrophic failure; testing would have tobe conducted to determine safe operational ranges.

ConclusionThe main challenges of the Eos II rocket were an under-

performing engine (and insufficient time to bring it up tofull power), and an oxidizer actuation and ground com-munications link that complicated the avionics system anddemanded intense time and effort to complete. While afunctional avionics system was ultimately obtained, thereis still significant work to be done in the way of improv-ing engine performance as detailed above. Overall, thecomprehensive design process taken to construct the EosII resulted in a by and large successful launch vehicle,whose strengths and weaknesses are known and ready tobe considered in future launch vehicles designed by UTATRocketry. In particular, the use of commercially availablesoftware for CFD was adequate when appropriate flow

models were used, whereas the inaccuracies of the in-house MATLAB Engine Performance Tool are known andamenable to revision. The emphasis on safety and lowcost exhibited by the Eos II should be supplemented witha focus on greater design simplicity and smoother inter-nal logistics, to facilitate ease of manufacture and rapidengine testing.

References[1] E. Doran, J. Dyer, K. Lohner, Z. Dunn, M. Marzona,

and E. Karlik, Peregrin Sounding Rocket. Stanford, CA:Stanford University, 2008.

[2] R. Humble, G. Henry, and W. Larson, Space PropulsionAnalysis and Design. Learning Solutions, 1995.

[3] A. Karabeyoglue, Hybrid Rocket Propulsion for FutureSpace Launch. Stanford, CA: Space Propulsion Group,2008.

[4] B. Genevieve, M. Brooks, P. Beaujardiere, andL. Roberts, Performance Modeling of a ParrafinWax/Nitrous Oxide Hybrid Motor. Orlando, FL: Amer-ican Institute for Aeronautics and Astronautics, 2011.

[5] R. Newlands, The Physics of Nitrous Oxide. Boston,MA: Boston University, 2011.

[6] Optical Research of Spray Development of E85 Fuel inHigh Pressure Gasoline Direct Injection, Amman, Jordan,2010.

[7] E. Toulson, Introduction to Combustion. East Lansing,MI: Michigan State University, 2014.

Supplementary TextAdditional systems were required to incorporate the

Bia II hybrid engine with a full rocket. Below are descrip-tions of key aspects of associated structures and avionicssubsystems, including airframe, bay connectors, recoverysubsystem, payload, wiring and connections, ground con-trol subsystem, power and control board, Flight ComputerA and Flight Computer B (Note: due to the supplementarynature of this section, figures have not been included andreaders are encouraged to contact the authors for moredetails).

AirframeTo reduce total weight, the Eos II main support struc-

ture was made from carbon fibre. Carbon fibre posedvarious difficulties for assembly and construction. Themain issue was radiative heat transfer to the carbon fi-bre in direct sunlight. Carbon fibre has high absorptivityand low reflectance levels, and hence can become quitehot in direct sunlight. 1D heat transfer calculations deter-mined the carbon fibre temperature to be 85C in expectedlaunch ambient conditions. At this temperature, the epoxywhich held the carbon fibre together would lose its struc-tural integrity, becoming about 70% as strong as when atroom temperature (Epoxy used for interior fixtures wasE120HP from Loctite). In order to mitigate the effects ofthis, Eos II was painted white. Calculations showed thiswould reduce the wall temperatures to near ambient lev-els, which was anticipated to be around 40C. Fabricationwas achieved with a 1

2 female mold made from Renshape470. The mold was CNC-milled to ensure its accuracy.A number of locating holes were placed on the mold inorder to allow for accurate placement of parts inside thestructure. Vacuum-bag layups were performed to createthe airframe.

Bay ConnectorsThe bay connector assemblies linked the various bays

of the Eos II rocket together. This assembly facilitatedquick assembly and disassembly, contained the requiredelectrical connections, and provided pressure sealing capa-bilities where required. The bay connector assembly alsoallowed for the fastening of launch lugs which connectedwith the launch rail. This assembly contained two maincomponents, the female connectors, which were epoxiedto the carbon fibre structure of each bay, and the maleassemblies, which link two adjacent females. The maleconnector assembly comprised 5 custom parts: 1 male con-nector, and 4 ledges. The ledges physically linked the maleconnector assembly to the female. The ledges were lo-cated to the male connector via dowels, and springs wereattached to the interior side of the ledges. The springs al-lowed for the ledges to be automatically pushed outward,allowing for a quick-connect assembly. This permitted thebays of the rocket to be assembled rapidly. When readyfor flight, screws on the outside of the bay were tightenedsuch that the ledges became locked in place. There werealso O-ring grooves, which provided pressure sealing ifdesired. The female bay connectors were epoxied to theends of each bay. The interior curvature of the femaleconnector was designed such that the ledges would lockin place when pushed into the female connector. The useof this three connector setup meant that each bay couldbe accessed from both sides, which greatly increased theease of assembly. Due to the high tolerance requirementsof this assembly, the male connectors and ledges wereall CNC manufactured. However, the female connectorswere not, owing to the lack of access to a CNC lathe tocreate the interior curvature. Instead, a custom lathing

tool was made to manually but accurately manufacturethe interior curvatures of the female connector.

Recovery SubsystemThe recovery system was comprised of a main and

drogue parachute, a tender descender, and several lengthsof high-strength rope. In order to actuate the recoverysystem, an explosive charge would be set off at apogee,separating the nose cone from the main rocket body. Thenose cone was held in with a 3o taper fit connection. Thetaper fit allowed for a secure connection before separation,but easy separation the charges were set off. Furthermore,the action of separating the nose cone would remove thedrogue chute from the recovery bay. The drogue was heldonto the nose cone via a tender descender, which wasthen fastened to the main structure. The tender descenderallowed for dual deployment, whereby the drogue wouldbe released at apogee, and the main chute would be re-leased at a lower altitude to reduce rocket travel distancesduring descent. Once the tender descender was released,the drogue would pull the main chute from the recoverybay.

PayloadEos II was designed to launch a 1U CubeSat, however

due to time constraints, the CubeSat could not be devel-oped without compromising design and testing of therocket itself. Instead, a 4.5kg metal weight was housedinside the nose cone, which would deploy at apogee.

Electronics and Avionics SystemThe electronics of the Eos II were divided into three dif-

ferent systems: the ground control system (GCS) and twoflight computers—Flight Computer A (FCA) and FlightComputer B (FCB). The GCS and the two flight computerswere connected via wired ground link, whereas the twoflight computers themselves were connected through alocal UART communication link. The purpose of the elec-tronics system was two-fold: (1) to attend to the needs ofthe Bia II engine ignition, oxidizer actuation, and teleme-try on the oxidizer tank while filling and (2) to implementan experimental algorithm for recovery actuation.

Ground Control SubsystemThe ground control system was comprised of an ig-

nition circuit and a ground link circuit attached to anArduino Mega 2560, which was in turn attached via USBto a computer. The computer itself was equipped withdaylight-readable screens, a durable Pelican case, andintegrated keyboard and mouse in order to increase itsrobustness for desert operation. The ignition circuit wasan implementation of a ‘firing circuit’, a custom modular-design solid-state pyrotechnic actuation circuit. The cir-cuit had two different MOSFETs that controlled currentflow, labeled ‘arm’ and ‘fire’. When the circuit was armed(the ‘arm’ FET is switched into saturation mode) the e-match attached in series would be energized to 9V, and

when the ‘fire’ FET was switched on also, sufficient cur-rent would flow in order to activate the e-match. Thecircuit was designed so that prior to the arming, therewould be no voltage applied to the e-match, and thereforeno risk of a short circuit. The reason behind choosingsolid-state construction was that relays (typically used forpyrotechnic actuation) are susceptible to launch vibrations.An additional feature of the firing circuit was continuitydetection across the e-match, which allowed for the de-tection of incorrectly attached e-matches. The circuit wasdesigned to be modular, and was implemented in such away that multiple channels of e-matches could be armedtogether but fired separately. Each channel could, in turn,fire three e-matches. There was a wired communicationlink between the ground station and the flight comput-ers. This was accomplished via an RS-485 connection ona CAT5e cable. Given RS-485’s design for long distancecommunication via differential driving, it was a robustchoice for the ground link which required up to a 100ftseparation between vehicle and ground station. On eachend of the ground link, the RS-485 was transformed intoa half-duplex UART link by an RS-485 transceiver. Onthe ground station side was placed am ADM483E, a 5Vslew-limited chip from Analog Devices. In addition, theRS-485 link was terminated using a 100Ω resistor, pre-venting reflections through the line by way of impedancematching with the characteristic impedance of the CAT5ecable. The half-duplex nature of RS-485 meant that onlyone side of the ground link (either the flight computersor the ground control system) could transmit at any onetime. Consequently, a custom protocol was designed toaccommodate this issue. In this protocol, the flight com-puters drove the communication. They either sent a datapacket (encapsulated via a header and footer) or sent atoken packet, to which the ground control system wouldrespond with either an acknowledge or an acknowledgewith a command attached (‘AckC’). In this fashion, theflight computers could convey state and telemetry infor-mation, while the ground station could convey commands.The token packages enabled the ground station to senda command at times that the flight computers did nothave any data to send, as well as established a ‘heartbeat’across the communication so that a disconnection couldbe instantly known.

Power and Control Board (P&C)The purpose of the Power and Control Board was to

distribute power to: (1) Flight Computer A (see section be-low for more information) Flight Computer B (see sectionbelow for more information) and the Recovery Firing Cir-cuit, (20 provide a way by which different circuits could beswitched on and off, and (3) power the LEDs to visualizedifferent states of the flight computers. In total, there werethree batteries that fed into the P&C board: a 2000mAhrechargeable Ni-MH battery pack, a 9V battery for recov-ery firing, and another 9V battery for the Raven3 com-mercial altimeter. The 12V battery was regulated downto the 3.3V operating voltage of the elect regulator, an

OKI-78-SR. This enabled the 12V circuits—the power forthe oxidiser actuation and the pressure transducers—tooperate on the same battery as all of the other circuits,reducing the number of batteries required for avionics.In total, the P&C board had five switches and four LEDs.These five switches controlled:

• Raven3 9V power

• FCA 3.3V power

• FCA 12V power

• FCB 3.3V power

• Recovery 9V power

The LEDs, on the other hand, indicated the power stateof the flight computers. Two green LEDs indicated thecommunication status of the link to the ground station,allowing an easy way to tell if the communication wasworking correctly. Two red LEDs indicated the safetystates of: (1) the 12V to FCA, which was used to drivethe actuation motor, and (2) the 9V source to the recoverywhich was used to deploy recovery.

Flight Computer A (FCA)Flight Computer A had several roles, mostly related to

pre-launch telemetry:

• Send oxidizer tank temperature and pressure datato the ground station

• Actuate the main oxidizer valve

• Send battery voltage data to the ground station

• Control the ignition sequence of the rocket

• Log flight data from FCB to a microSD card

The microcontroller on FCA, common across both theflight computers, was an ATmega1284P. This microcon-troller was socketed as a DIP package, allowing easy re-moval, and, if needed, replacement during programmingand debugging. This microcontroller was programmedwith a non-standard version of the Arduino bootloader inorder to allow programming over an FTDI USB to UARTadaptor. This gave easy access to all the libraries builtaround the open Arduino platform to leverage a fasterprogramming and testing cycle. The pressure transducerin Eos II was a 1000 psi model that fed its data outputin the form of a current between 4-20mA full scale (Thepressure transducer used is the Swagelok PTI-S-NG1000-12AQ). In order to convert this to a voltage that the micro-controller’s onboard analog-to-digital-converter (ADC)can read, a noninverting amplifier op-amp circuit wasused. However, one potential issue with this circuit wasits temperature stability, as the resistors used in this ana-log circuit were temperature dependent and therefore

the readings could change with changing ambient tem-peratures. The effect of this was not determined. Thethermocouple amplifier on FCA was a MAX31855, whichinterfaced with the microcontroller over the SPI proto-col. This particular chip used cold-junction compensation,enabling the use of a thermocouple without a constanttemperature junction on the other side. As well, since thethermocouple amplifier also had an internal temperaturesensor, it was used to sense the instantaneous temperatureof the interior of the avionics bay. The oxidizer actuationsystem was composed of an H-bridge—an off-the-shelfTLE5206—and the relevant pins used to interface withthe encoder and limit switches of the oxidizer actuationhousing. The encoder pin was fed into the microcon-troller’s timer/counter input, which enabled counting ofthe square-wave pulses of the encoder without the useof interrupts. The switches were fed directly into GPIOpins of the microcontroller. Through testing, it was foundthat the microcontroller was not very consistent at read-ing the states of the switches. When observed using anoscilloscope, it was found that the switch lines had a con-siderable amount of noise. This was attributed to thefact that the high frequency pulses of the square-waveencoder were adjacent to the switch lines. However, asolution was found in delaying the switch polling code,for unknown reasons. The RS-485 transceiver chip on thisflight computer was the ADM3072E, a 3.3V chip similar tothe one used in the ground control system. It was directlyconnected to the same UART port as the FTDI header forprogramming the microcontroller, which created someissues in programming the ATmega with the RS-485 lineactive. This was worked around by using a softwaretimeout before activating the transceiver. Again, the RS-485 differential lines were terminated with 100Ω resistors.Communication to the microSD card was completed usingan SPI bus. While this meant that the microSD card wouldrun at a slower speed than normal, it ensured easy pro-gramming for the microcontroller. It was included in FCAinstead of FCB because in past experience, writing to theSD card took a long time in contrast to the other parts ofthe main loop of the microcontroller. Separating the roleof gathering data and deploying recovery from recordingdata meant that the deployment algorithm could run at ahigher rate. The battery voltage data was obtained usinga voltage divider for each battery, feeding into the built-inADC on the microcontroller and relayed to the groundstation for monitoring.

Flight Computer B (FCB) and Recovery Firing Circuit

Flight Computer B had one role: to deploy recoverywith a custom algorithm. To that end, it had the bareminimum of peripherals attached to the ATmega1284P:Two-channel implementation of the firing circuit Adafruit10DOF IMU with accelerometer, gyroscope, barometer,and magnetometer RS-485 transceiver UART link to FCAto log data to the SD card An IMU was selected instead ofindividual sensors because of the relative cost-effectiveness

of the entire sensor suite versus buying each sensor sep-arately on a breakout board. Each sensor could not behand soldered, as they were often packaged in a no-leadsurface mount package. The RS-485 transceiver was thesame chip as in FCA, an ADM3072E. A Kalman Filter wasused in FCB to aid with state estimation as a part of therecovery system, and the design was based upon work bySchultz [4]. Some simplifying assumptions were made, in-cluding that the state transition matrix F, the model covari-ance matrix Q, and the sensor covariance matrix R wereconstant. Using all of these assumptions as well as theassumption that the acceleration is constant, which wasreasonable during coasting, allowed the construction of aconstant Kalman gain matrix. This allows Kalman gainsto be pre-computed given barometer variance, accelerom-eter variance, model variance, and a value to specify thenumber of iterations to use in determining the values forthe Kalman gain. An algorithm took the Kalman filteredvalues from the sensor and checked whether or not thevelocity is greater than 1 m/s downward. If it was greaterthan that, then the drogue parachute would be deployed.The main parachute would then be deployed at 1500ftabove ground level.

Feasibility of Single-Link Inverted Pendulum Modelsfor Human Standing PostureKai Lon Fok1 and Kei Masani2,3

Human standing posture can be modeled as a single-link inverted pendulum (SIP) model rotating aboutthe ankle joint or a double-link inverted pendulum (DIP) model rotating about the ankle and hip joints.Here we investigated during standing (1) the “rigidness” of SIP, that is, how much the length of the centreof mass (COM) (i.e., distance from the ankle joint to the COM) changes, (2) the degree in which the anklejoint motion represents the SIP model, and (3) to what degree the ankle joint torque can represent the SIPmodel. Subjects were instructed to stand quietly or to sway voluntarily with their arms crossed againsttheir chest. The 3D body kinematics was recorded by a motion capture system and the reaction force wasrecorded by a dual force plate. To determine the rigidness of SIP we analyzed the percent changes of theCOM length and found that the length changed less than 0.2%. Furthermore, we analyzed the correlationbetween the ankle joint and SIP COM angle, along with their fluctuations and found that the differencesbetween the fluctuations were negligible. To determine the accuracy of using the ankle joint torque torepresent the SIP model we calculated the ankle joint torque using both the SIP and DIP models, whilecomparing them to the ankle joint calculated using kinetic information from the force plates. It was foundthat all three torques had a correlation greater than 90% and the differences were statistically negligible.We conclude that the ankle joint kinematics and kinetics represents the COM behaviour during quietstanding and during voluntary sway with small errors, suggesting that simplification of standing postureusing the SIP model is feasible when unperturbed standing such as quiet standing and voluntary swayingis investigated, as long as the described error is considered.

IntroductionCoordination of joints during standing is usually clas-

sified into two strategies: ankle strategy and hip strategy[1]. When the postural sway is small, the ankle joint isdominantly used, which is called the ankle strategy. Whenthe postural sway becomes larger, the hip joint tends tocontribute to the control of the body, which is called thehip strategy. During quiet standing (i.e., standing withoutany other limb movements), as the postural sway is small,it was believed that the ankle strategy is dominantly used.Therefore, the upright posture during quiet standing wasoften modeled as a single-link inverted pendulum (SIP)[2, 3, 4, 5, 6, 7, 8]. The SIP model represents the humanbody as one rigid segment rotating about the ankle joint.The advantage of the SIP model is the simplification ofcontrol problems, such as facilitating investigations onthe control mechanism of the ankle join which primar-ily balances the equilibrium of human body against the

1Division of Engineering Science, University of Toronto, Toronto, ON2Institute of Biomaterials and Biomedical Engineering, University of

Toronto, Toronto, ON3Toronto Rehabilitation Institute, University Health Network,

Toronto, ON

environment [2, 3, 4, 5, 6, 7, 8], and facilitating the de-velopment of assistive technology for standing balance[9, 10, 11].

However, in practice, an upright standing posture wouldnever be a rigid single-link segment. For example, Ac-cornero et al. [12] revealed that the hip joint action is notignorable even during quiet standing, and Aramaki et al.[13] demonstrated that the ankle and hip joint actions arereciprocal resulting in a reduced body acceleration. Fur-thermore, Sasagawa et al. [14] showed that the hip jointacceleration can induce acceleration in the ankle joint. Allof these suggest that the hip joint action is considerableand can affect the ankle joint action, leading to the require-ment of a double-link inverted pendulum (DIP) modelfor standing posture [14] where the body is assumed tobe two rigid segments rotating about the ankle and hipjoints.

Thus, even though the SIP model has been used inmany previous studies because of its various advantages,the DIP model is a more accurate representation of uprightstanding posture. However, there is no study that quanti-fied the amount of error in the SIP model, which shouldbe identified when we use the SIP model. Therefore, weaimed to quantify the amount of error in the SIP model in

three aspects: (1) the “rigidness” of the SIP model, that is,how much the distance from the ankle joint to the COM(length of the COM) changes, (2) the degree in which theankle joint motion represents the SIP model, and (3) towhat degree the ankle joint torque can represent the SIPmodel.

MethodsSubjects

Ten healthy males (age 19.9 ± 2.3 years; height 174± 5.8 cm; weight 69 ± 8.7 kg) participated in this study.They had no medical history or signs of neurological dis-orders. All subjects gave written informed consent toparticipate in the study, and the experimental procedureswere approved by the local ethics committee.

ProcedureSubjects performed two separate tasks during this ex-

periment: quiet standing (QS) and voluntary sway (VS).For each task subjects stood barefoot with their armscrossed against their chest with their eyes opened. DuringQS, the subjects were instructed to stand relaxed and to fo-cus on a marker placed eye level 2 meters in front of themfor 125 seconds. The VS tasks involved eight trials, all con-ducted while the eyes were open. Subjects were informedof eight directions (numbered one through eight) at thebeginning of the task. Subjects were then instructed ofwhich direction to gently sway towards when prompted(every 10 seconds, for 125 seconds) for each trial. The3D body kinematics was recorded by a motion capturesystem (Cortex V3.1.101, Motion Analysis Corp., SantaRosa, USA) composed of six infrared cameras in a cir-cular arrangement. Twenty-nine reflective markers wereaffixed to the body according to the modified Helen Hayesskeletal outline [15]. The reaction force was recorded by adual force plate (Accu Sway ACS-DUAL, Advanced Me-chanical Technology, Watertown, USA). Data for the bodykinematics and the reaction force were synchronouslyrecorded with a sampling frequency of 200 Hz and 2 kHz,respectively.

AnalysisKinematics

In this paper, we focus only on the anterior-posteriordirection where the postural sway is more prominent dur-ing standing [4]. The time series of each joint coordinatewas digitally smoothed with a zero-phase lag, second-order low-pass Butterworth filter with a cutoff frequencyof 2.0 Hz [7]. Based on the smoothed joint coordinates, wecomputed the lengths of the two segments (i.e., lower andupper body segments). The length of lower body segmentwas defined as the distance between the ankle and hipjoints. The length of upper body segment was definedas the distance between the hip joints and the top of the

head. We calculated the COM in two ways: one was calcu-lated directly from all the segments in the whole body (SIPCOM) and the other was calculated as a weighted averageof the lower and upper body segment COMs (DIP COM).The angles of the lower and upper body segments weredefined as the ankle joint angle (θa) and hip joint angle(θh) respectively, while the angle of the segment betweenthe ankle and the COM was defined as the COM angle(θCOM) for DIP COM and SIP COM separately (see Figure1).

Fig. 1. Schematic representation of calculated parameters.COM denotes the center of mass of the whole body. θaand θh denote angles of the lower and upper bodysegments respectively. θCOM denotes the angle of COM.

During each standing task, the length of each segment(lower/upper body segments and DIP/SIP COM) changed.We quantified the amount of change using the root meansquare (RMS) for each segment for each task. Accordingto Aramaki et al. [13], a reciprocal action between upperand lower body segments was expected. This resulted ina possibility that the ankle joint angle would not repre-sent the SIP COM angle. To investigate this, we analyzedthe correlation between the ankle joint and the SIP COMangles, and also compared the amount of angular fluctua-tions using RMS.

Calculation of ankle and hip torqueThe ankle joint torque was calculated in three ways: (1)

using force component and the centre of pressure (COP)from the force plate(AnkTQFP) as shown in equation 1[7].

(AnkTQ)FP = FvPx (1)

Where Fv is the vertical reaction force component (ex-cluding the feet) and Px is the horizontal distance betweenthe ankle joint and the COP, (2) using kinematics with DIP

model (AnkTQDIP), and (3) using kinematics with SIPmodel (AnkTQSIP). AnkTQDIP was calculated accordingto equation 2 [14].

(AnkTQ)DIP = θa[Θ1 + Θ2 + m1r21

+ m2r22 + m2l2

1 + 2m2l1r2cosθh]

+ θh[Θ2 + m2r22 + m2l1r2cosθh]

+ θ2h[−m2l1r2sinθh]

+ θa θh[−2m2l1r2sinθh]

+ g[m1r1cosθa + m2l1cosθa

+ m2r2cos(θa + θh)]

Here, Θ represents the moment of inertia about thecentre of gravity, r1,2 is the distance to the COM of eachsegment from the distal joint of each segment; m1,2 is themass of each segment with subscripts 1 and 2 correspond-ing to segments 1 and 2 respectively. Note that Θ1 and m1are values of both legs combined. Also Θ, r1,2 and m1,2were estimated using the standard anthropometric tablesof Winter [16]. AnkTQSIP was calculated using the COMkinematics (i.e., the angular acceleration of COM) and itsmass according to equation 3.

(AnkTQ)SIP = MR2θCOM − grsinθCOM (3)

Where M is the mass of the subject, R is the lengthof the COM, g is the gravitational constant and θCOM isthe angle of COM. θCOM is the angular acceleration andwas calculated by differentiating the angular displace-ment data with a three-point central difference formula[16].We examined the correlation between AnkTQDIP andAnkTQSIP, and AnkTQFP to investigate how close theirfluctuations were. We compared the fluctuation amountsof the three using RMS. One way analysis of variance wasused with a significance level of 95% for each statisticalcomparison.

ResultsKinematics

Figures 2A and 2B show time series of the percentchanges of the segmental length for each segment (up-per/lower body segments and DIP/SIP COM) for QS (A)and VS (B). Figures 2C and 2D show the group results.The lower body segment and DIP/SIP COM fluctuatedless than 0.08% during QS and 0.2% during VS, whileupper body segment fluctuated less than 0.4% during QSand 0.6% during VS. Thus, the percentage of change was,overall, very small, while the large change of the upperbody length must be a result of the head movements ofthe subject.

Fig. 2. Percent changes of segmental length for eachsegment including upper/lower body segments andDIP/SIP COM for QS (A) and VS (B). For the purpose ofvisualizing each fluctuation, only 15 seconds offluctuation are presented here. The group means andstandard deviations for each percent change of segmentallength are shown for QS (C) and VS (D).

Figure 3A shows an example relationship between theankle joint angle and the COM angle. They showed astrong linear correlation, suggesting that they fluctuatedsynchronously. Figure 3B shows the group average of cor-relation coefficient between the ankle joint angle and theCOM angle for QS and VS (0.868 ± 0.17 and 0.941 ± 0.01).The results suggest that they were highly correlated in allsubjects in either task. Figure 3C shows the comparison offluctuation amount between the ankle joint angle and theCOM angle (0.363± 1.3 for θa and 0.300± 0.11 for θCOM

in QS, and 1.45± 0.20 for θa and 1.40± 0.18 for θCOM inVS). There were no significant differences between themfor QS and VS. Figure 3D shows the percentage error be-tween the ankle joint angle and COM angle which wereless than 2%.

Fig. 3. Example relationship between the ankle jointangle and the COM angle for QS(A), group average ofcorrelation coefficient between the ankle joint angle andthe COM angle for QS and VS (B) and comparison offluctuation amount between the ankle joint angle and theCOM angle for QS and VS (C), error bar representationsof the error between the ankle joint angle and SIP COMangle for QS and VS (D).

KineticsFigure 4A shows an example time series of the ankle

joint torque for a QS trial. The torque fluctuations wereidentical among the three, however a vertical shift ex-isted between AnkTQFP and AnkTQDIP and AnkTQSIP.This vertical shift was a result of a technical misalignmentbetween the force plate and the kinematic coordinates.Figure 4B shows the group result of correlation coeffi-cients between AnkTQFP and AnkTQDIP and AnkTQSIP.There were very high correlations between AnkTQFP andAnkTQDIP and AnkTQSIP for both QS and VS. Figure4C shows the group results of torque fluctuation amount.There were no significant differences in the torque fluctua-tions among AnkTQFP, AnkTQDIP and AnkTQSIP. Figure4D shows the percentage error between AnkTQDIP andAnkTQSIP and AnkTQFP (8.6 ± 5.4% and 15.2 ± 11.2%for AnkTQSIP and AnkTQDIP for QS, and 19.6 ± 11.6%and 27.4 ± 17% for AnkTQSIP and AnkTQDIP, for VS re-spectively). Although the errors were considerable (9%and 15% for AnkTQSIP and AnkTQDIP for QS, and 20%and 27% for AnkTQSIP and AnkTQDIP, for VS respec-tively), there were no statistical differences between thethree torques.

DiscussionWe demonstrated that the change of length between

the ankle joint to the COM was small, i.e., less than 0.08%during QS and less than 0.2% even during VS (Figure2C and 2D). These results suggest that the SIP model as-

Fig. 4. Example time series for a QS trial displaying thevarious ankle torques (A), group average of correlationcoefficient between the AnkTQDIP to AnkTQFP andAnkTQSIP to AnkTQFQ respectively for QS and VS (B)and a comparison of the fluctuation amount betweenAnkTQDIP, AnkTQSIP and AnkTQFP for QS and VS (C),error bar representations of the error betweenAnkTQSIP/AnkTQDIP and AnkTQFP for QS and VS (D).

sumes the body to be a rigid segment as long as thesetiny changes are considered. We also found that the ankleangle was highly correlated with the COM angle dur-ing QS and even during VS (Figure 3B) and the amountof fluctuation was not statistically different between theankle joint and the COM angles. These results suggestthat the ankle joint angle represents the COM angle with asmall error regardless of the reciprocal behaviour betweenupper and lower body segments [13]. Consequently theresults demonstrate the feasibility of the SIP model, as thehip joint movement did not greatly influence the COMangle.

Sasagawa et al. [14] demonstrated that the accelera-tion of the upper body segment can affect the angularacceleration of the lower body segment, i.e. induced ac-celeration, which can affect the joint torque calculation.Despite this condition, the ankle torque calculated usingkinematic information with SIP (i.e., AnkTQSIP) and DIP(i.e., AnkTQDIP) models were almost identical, and theankle torque fluctuations were identical to the one calcu-lated based on the force plate data (AnkTQFP) (Figure 4C).The errors of AnkTQSIP and AnkTQDIP from AnkTQFPwere considerable due to the methodological differences,the two models provided equivalent accuracy (Figure 4D).This result suggests that, even with induced accelerationaffecting the ankle joint torque; the ankle joint torque cal-culated based on the SIP model is acceptable. Thus, wedemonstrated that (1) the change of the Com length isless than 0.08% during QS and less than 0.2% even duringVS, (2) the ankle joint motion sufficiently represents theSIP model, and (3) the ankle joint torque also sufficientlyrepresents the SIP model. As a result, the SIP model cancontinue to be used to simplified control problems, greatlyreducing the complexity of studies and technologies re-quiring an understanding of the human body’s balancesystem. Future experiments can be conducted analyzingthe body’s use of its muscles in human standing. The SIP

model then can be used to help develop a controller toassist in human standing, with the information on muscleactivity during human standing.

ConclusionWe experimentally demonstrated that the ankle joint

kinematics and kinetics represent the COM behaviourduring quiet standing and during voluntary sway, whileconsidering the reciprocal behaviour between lower andupper body segments [13] and the induced acceleration[14]. These results suggest that the simplification of stand-ing posture using the SIP model is feasible when unper-turbed standing such as quiet standing and voluntaryswaying is investigated, as long as the described error isconsidered.

AcknowledgmentsThe authors acknowledge the support of the Toronto

Rehabilitation Institute-University Health Network whichreceives funding under the Provincial Rehabilitation Re-search Program from the Ministry of Health and Long-Term Care in Ontario. The authors would also like tothank Eric Ma and Daniel Chung for their technical assis-tance during data collection.

References[1] F. B. Horak and J. M. MacPherson, “Postural orienta-

tion and equilibrium,” Bethesda: American PhysiologySociety, pp. 225–292, 1996.

[2] R. C. Fitzpatrick, J. L. Taylor, and D. I. McCloskey,“Ankle stiffness of standing humans in response toimperceptible perturbation: reflex and task-depedentcomponents,” Journal of Physiology, vol. 454, pp. 533–547, 1992.

[3] I. D. Loram, C. N. Maganaris, and M. Lakie, “Active,non-spring-like muscle movement in human postu-ral sway: how might paradoxical changes in musclelength be produced?” Journal of Physiology, vol. 564,pp. 281–293, 2005.

[4] K. Masani, M. R. Popovic, K. Nakazaka, M. Kouzaki,and D. Nozaki, “Importance of body sway veloc-ity information in controlling ankle extensor activi-ties during quiet stance,” Journal of Neurophysiology,vol. 90, pp. 3774–3782, 2003.

[5] P. G. Morasso and M. Schieppati, “Can muscle stiff-ness alone stabilize upright standing?” Journal ofNeurophysiology, vol. 82, pp. 1622–1626, 1999.

[6] R. Peterka, “Postural control model interpretation ofstabilogram diffusion analysis,” Biological Cybernetics,vol. 82, pp. 335–343, 2000.

[7] D. A. Winter, A. Patla, M. I. F. Prince, and K. Gielo-Perczak, “Stiffness control of balance in quiet stand-ing,” Journal of Neurophysiology, vol. 80, pp. 1211–1221, 1998.

[8] K. Masani, A. H. Vette, N. Kawashima, andM. Popovic, “Neuromusculoskeletal torque-generation process has a large destabilizing effect onthe control mechanism of quiet standing,” Journal ofNeurophysiology, vol. 100, pp. 1465–1475, 2007.

[9] K. Masani, A. H. Vette, and M. R. Popovic, “Con-trolling balance during quiet standing: proportionaland derivative controller generates preceding motorcommand to body sway position observed in experi-ments,” Gait and Posture, vol. 23, pp. 164–172, 2006.

[10] A. H. Vette, K. Masani, and M. R. Popovic, “Imple-mentation of a physiologically identified pd feedbackcontroller for regulating the active ankle torque dur-ing quiet stance,” IEEE Transcations on Neural Systemsand Rehabilitation Engineering: a Publication of the IEEEEngineering in Medicine and Biology Society, vol. 15, pp.235–243, 2007.

[11] A. H. Vette, K. Masani, J. Y. Kim, and M. R.Popovic, “Closed-loop control of functional electricalstimulation-assisted arm-free standing in individualswith spinal cord injury: a feasibility study,” Neuro-modulation, vol. 12, pp. 22–32, 2009.

[12] N. Accornero, M. Capozza, S. Rinalduzzi, and G. W.Manfredi, “Clinical multisegmental posturography:age-related changes in stance control,” Electroen-cephalography and Clinical Neurophysiology, vol. 105,pp. 213–219, 1997.

[13] Y. Aramaki, K. M. D. Nozaki, T. Sato, K. Nakazawa,and H. Yano, “Reciprocal angular acceleration of an-kle and hip joints during quiet standing in humans,”Experimental Brain Research, vol. 136, pp. 463–473,2001.

[14] S. Sasagawa, M. Shinya, and K. Nakazawa, “Inter-joint dynamic interaction during constrained humanquiet standing examined by induced accelerationanalysis,” Journal of Neurophysiology, vol. 111, pp. 313–322, 2014.

[15] M. P. Kadaba, H. K. Ramakrishnan, and M. E.Wooten, “Measurement of lower kinematics duringlevel walking,” Journal of Orthopaedic Research, vol. 8,pp. 383–392, 1990.

[16] D. Winter, Biomechanics and Motor Control of HumanMovement. New York: Wiley, 1990.

Part II: News Articles

Legacy Uranium Mining Threatens Water SupplySharon Mandair1

Study of water, soil, and vegetation in the farms surrounding old uranium mines in Portugal reveals highlevels of trace radioactive material.

Environmental chemist, Dr. Fernando P. Carvalho, spe-cializes in radiology, specifically looking at the the legacyof uranium mining in his home country, Portugal. Withhis team of researchers at the University of Lisbon, hefound radioactivity in garden vegetables in Portugal’scountryside.

The team studied three towns in the north of the coun-try where uranium mining had at one time been part ofthe local economy: Nelas, Mangualde, and Sabugal. Thesetowns are home to vast agricultural land, including or-chards, vineyards and home gardens. The radionuclideswere found by taking measurements of water, soil, fruitand vegetables both upstream and downstream of the oldmines, which had continued to release waste into the localwater courses even after their closure [1].

To no surprise, high levels were found of a variety ofradionuclides, including a number of variants of uraniumand radium. The highest result found was 2.8 becquerelsper kilogram of Uranium-238 in a water cress in Sabugal[1]. This is equivalent to 23 micrograms of radiation in a100 grams water cress.

To put this into perspective, the World Health Organi-zation recommends a tolerable daily intake (TDI) of 0.6micrograms per kilogram of body mass [2]. This amountcan be ingested safely each day over a life time. So for a60 kilogram person, according to this model, they shouldbe able to eat 36 micrograms a day for over a lifetimewith no side effects. Most people are exposed to about 3micrograms a day from food [2].

Transport MechanismsBesides these findings, the study made some interest-

ing conclusions that help us to understand how radionu-clides find their way into crops.

Firstly, and what might seem most obvious, the furtherdownstream from a mine crops were grown, the moreradioactivity was found. This is shown most clearly inNelas, which is illustrated in the diagram below. Thisshows the build up of the material downstream, comparedto upstream of the mine, where very little is found.

Secondly, they found that plants take up radionuclides

1Department of Civil Engineering, University of Toronto, Toronto,ON

Fig. 1. The three study sites: Nelas, Mangualde, andSabugal.

from the water they are fed, not the soil. This was madeclear by comparing apples fed by contaminated surfacewater, to apples fed with water irrigated from an uncon-taminated groundwater source. Both were in the clusterdownstream of the east mine in Mangualde, but the ap-ples fed with uncontaminated water had a fifth as muchradiation!

These measured findings of radiation are of course im-portant for the affected communities as they are able toidentify whether there is a significant health risk fromeating their home grown food. Moreover, the conclu-sions drawn from the evidence (how radionuclides get

into crops) gives the communities some confidence thatthey can avoid the risks associated with the contaminatedwater by switching sources.

Portugal’s Mining HistoryUranium and radium mines were numerous in Portu-

gal throughout the 1900s. In northern Portugal alone, thearea of study, there were sixty mining operations of thissort, including those in Nelas, Mangualde and Sabugal.The industry was accelerated during the Second WorldWar when supplies were tapped for the creation of mil-itary weapons, and later fuelled nuclear energy gener-ation [3]. In 2001, the market value of these resourcesplummeted, and mines across western Europe stoppedoperations.

Fig. 2. Amount of radiation in lettuce along the stream inNelas.

“[G]uidelines establishing radioactivity limits for. . . water and soils totally lack in national legisla-tion, which renders advice and regulatory controlof water and soil use very difficult.

”- Fernando P. Carvalho

The three mining towns from the study are quite typical.Mining operations often set up near towns to benefit fromtransportation infrastructure and lodgings for workers [3].

Fig. 3. Radionuclides move with the water taken up bythe plants.

A report from the International Atomic Energy Agency(IAEA), reviewing the environmental consequences ofthis mining industry, explains “the practice at almost ev-ery uranium processing plant is to dispose of the tailingsat the nearest convenient place" [3]. Unfortunately, thismeant waste disposal from mining processes was oftenlocated near these towns, “where the risk of disturbanceand exposure to the general population was highest” [3].

Reflecting on his research and the industry’s history,Dr. Carvalho gave his lessons learnt at and IAEA con-ference in 2009. He stresses there was a lack of foresightas the mining industry had not planned for post miningrestoration, and as a result there is limited funding forit [4]. So far, rehabilitation has been costly, but environ-mental remediation of these former mining sites is needed[4].

He also acknowledged in his research paper anothermajor hurdle to come: establishing guidelines. “Once allthe mine water discharges [is] treated, the quality of sur-face waters will improve and the transfer of radionuclidesto vegetables in agriculture plots might be prevented moreefficiently. Nevertheless, guidelines establishing radioac-tivity limits for several types of water, including irrigationwater and soils totally lack in national legislation, whichrenders advice and regulatory control of water and soiluse very difficult” [1].

Hope for RemediationDuring Dr. Carvalho’s study, the waste from the Urgeir-

iça mine, one of the two in Nelas, was addressed and hewas able to compare the effects before and after. Theresults show” radioactive discharges into the stream at

Urgeiriça have decreased after implementation of wastewater treatment," says Carvalho in his report [1].

The IAEA advocates for a strategic approach to ad-dressing the mining legacy. It starts with regulationsand guidelines, followed by remediation, monitoring andstewardship [3]. In the case of Urgeiriça, remediation wasdone by burying the tailings to isolate the contaminantsfrom surrounding environment [4]. The waste water isalso treated with neutralizing and precipitation reactions[1].

Sofia Barbosa, a geological engineer at EnvironmentalDesign & Management Ltd.(EDM) says in reference tothe heavily engineered remediation done at Urgeiriça,“these remediation works are too expensive. We have totry bioremediation approaches that have more naturalsolutions” [5].

At Aveiro University, also in Portugal, a different teamof researchers are looking into an innovative approachto remediation. Their goal is to adapt local plants andtrees to absorb radioactive contaminants in their roots,taking them out of the soil and water [5]. They hopethis technology will offer an affordable solution to theenvironmental legacy of the uranium mining industry inPortugal and other countries faced with similar problems[5].

References[1] Radioactivity in Soils and Vegetables from Uranium Min-

ing Regions, vol. 8, 2014.

[2] South and W. D. H. Authority, “Uranium in drinkingwater: Fact sheet,” Nation Health Service, January2001.

[3] Environmental Contamination from Uranium ProductionFacilities and their Remediation. Internations AtomicEnergy Agency, February 2004.

[4] F. P. Carvalhoa, “Environmental remediation and ra-dioactivity monitoring of uranium mining legacy inportugal,” Presentation, 2009.

[5] NATO. (2011, February) Using plants to removeuranium pollution from the portuguese soil. [Online].Available: https://www.youtube.com/watch?v=tRkQY-LCDIc

Why Does the Same iPhone Come Out Every Year?Sam Osia1

University of Toronto engineers are breaking new barriers to develop more advanced electronics.

The rapid advancement in microelectronic processorshas led to significant growth in everyday electronic de-vices. Not only are these devices exponentially more pow-erful with more features and capabilities, they are alsosmaller in size and lighter in weight.

Why is it then, that for the past few years, the newversions of cellphones have been the same as the previouswith a more glorified case and new number in front of it?How come all each company has to advertise about is anextra megapixel on the camera or a new swipe feature?Why is it that the newest generation of smart phones havealmost identical processing powers as their past four tofive generations?

While the research and development of the micropro-cessors continues with great success, there is another bar-rier hindering manufacturers from advancing further. Theincreased power in the processors is causing more heat tobe generated, requiring the need for bigger heat sinks andventilations to cool the system which prevents reducingthe size and weight of the device.

However, what if the case of the device did not trapthe heat in? What if it could allow the heat generatedby the processor through itself? Professor Naguib andhis team from the University of Toronto are working to-wards developing a new material to dissipate the heatwhile maintaining the same mechanical properties of theprevious casings such as impact resistance and density. Inorder to dissipate the heat, the material must have a highthermal conductivity.

The newly developed material is a composite with apolymeric matrix and a ceramic filler. This is essentiallylike a chocolate chip cookie with the cookie dough beingthe polymer and the chocolates being the ceramic. Thedough of the cookie, is soft and ductile while the choco-late is hard and brittle. When trying to bite the cookie,a plain cookie and a chocolate chip cookie will behavesimilarly; however, the chocolate chip cookie will havethe taste of the chocolate. Thus, it can be seen that the ma-jor mechanical properties are determined by the base ormatrix of the material while other properties such as taste,or thermal conductivity can be altered by the addition ofother materials such as chocolates or highly conductiveceramics in the respective order. In this case, when try-

1Department of Mechanical and Industrial Engineering, Universityof Toronto, Toronto, ON

ing to alter the thermal conductivity, the most efficientmethod is creating a path from one side of the polymer tothe other. This is similar to poking holes through a potatoor burger to allow the heat to travel across it and cook itmore evenly.

By using this method, they have managed to improvethe thermal conductivity of the original polymer by afactor of over 100. This ground-breaking method of syn-thesizing materials can finally break the plateau that theelectronic industry is in, specifically in terms of smartphones and laptops.

Another important aspect of this research which is con-sidered an advancement in the field of electronic pack-aging is the usage of biodegradable polymers. We allhave that one friend that always has the newest genera-tion of cellphones and tablets. This means that the rate atwhich electronics are being disposed of has increased andrecyclability of the material is now an important factor.

Therefore, from every direction, the future of this devel-opment seems bright. Professor Naguib, the lead profes-sor in this project says, “In brief, the future of this researchis to develop new manufacturing technology for electronicmaterials using thermal management composites. Thesematerials are lighter in weight, easy to manufacture, lowcost and recyclable.”

The news on when we can see this technology be im-plemented is yet to be announced as the research papersregarding the project have been accepted, but the officialpublication is pending. Furthermore, Professor Naguibhas announced that the names of the companies workingwith this project are to remain confidential for the timebeing.

However, despite not knowing the exact timeline, thereis finally a hope for technology enthusiasts that this in-dustry will be able to break through the rot and advancefurther in the near future.

The Growing Field of Personalized Health CareShendu Ma1

“Baymax,” the personal health scanner in the movie Big Hero 6 is possible in the future with researchachievement at the University of Toronto (UofT).

Impressed by the idea of personal heath assistant robotlike “Baymax” in the movie Big Hero 6? Concerned aboutyour health but tired of medical imaging and tests like theCT and the X-ray? A revolutionary biomedical imagingtechnology, frequency-domain photothermalacoustic (FD-PTA) imaging, developed by Prof. Andreas Mandelis’research team from the University of Toronto (UofT), haslept a big step from reality to science fiction.

Contemporary medical imaging methods are basedon diffusive waves - electromagnetic waves (X-ray) andultrasonic waves (B ultrasound) - that are capable of pen-etrating human tissues, especially skin and muscle. Theyare either harmful to human body or of low resolution[1]. Pioneers in PTA imaging field have confirmed thefeasibility of using pulsed lasers as a signal source fortissue imaging to avoid radiation from the X-ray. You canimagine conducting experiments of measuring the depthof a lake by determining the time needed for the ray totravel between the surface and the bottom of the lake. Youwon’t be able to sense the reflected laser with your eyesas it consumes energy on its way through the lake andbecame so weak that it’s invisible. You want to have a“receiver” (i.e. the transducer with wide band coverage)that can capture this weak reflected signal for you. Youalso want this “receiver” to calculate the travelling timefor you since the laser travels at the speed of light. To getthe time (usually hundredth of second), you won’t be ableto use a stopwatch since you can’t react quickly enough.Lastly, you want to maximize the power of the laser.

These difficulties you will experience are exactly prob-lems in contemporary research. The transducer (“receiver”)that works at wide bandwidth (0-100 Hz) is difficult toproduce [2]. The high peak power required by such exper-iments increase the potential damage brought by the lasersource to the tissue. The environment of human tissue ismuch more complicated than that of the lake. Most impor-tantly, the goal is to image the tissue, i.e. map the bottomof the lake in our analogy experiment; it is impossible torepeat this single experiment indefinitely.

“Why can’t we change the quantity we measure fromtime to energy loss (frequency change)?” you might askthe same question as Prof. Mandelis asked in the proposal

1Department of Mechanical and Industrial Engineering, Universityof Toronto, Toronto, ON

of his new approach. Instead of measuring the travellingtime (time domain), transducers are more sensitive to en-ergy change (frequency domain). Moreover, frequencydomain calculation is more straightforward than time do-main in correlation between waves and media it travelsthrough. Prof. Mandelis and his team structured the math-ematical model of this frequency domain PTA measure-ment. Experiments validate that by measuring frequencyresponse rather than time response, SNR is significantlyincreased, a much higher resolution of PTA imaging isachieved by this approach. The team was able to applythe theoretical results to image blood vessel under skintissues ex vivo in preliminary application experiment.

“For example, when a tumor starts to grow, it is ac-companied by the growth of new blood vessels,” Prof.Mandelis explained. “Lasers can spot these blood vesselsearlier than ultrasound machines because blood absorbslight differently. This allows for early detection of tumors.”Despite such great potential of this achievement, Prof.Mandelis’ research was not well known to the public untilhe won the Killam Prize last year. The Killam Prize is oneof Canada’s most prestigious scholarly awards, recognizesoutstanding research by scholars with a $100,000 prize [3].“We go where no light has gone before,” said Mandelis[4]. “In the past decade, IBBME (Institute of Biomaterialsand Biomedical Engineering) at UofT has concentratedits resources on tissue engineering and clinical engineer-ing. Not too much attention was on biomedical imaging.Winning the Killam Prize reminds us to recognize the con-tribution of and allocate more resources to the researchin such field,” explained by Prof. Craig Simmons, vicedirector of the IBBME.

You may wrongly think it will ma to finally use thistechnology in daily life. Prof. Mandelis has alreadyfounded the Quantum Dental Technologies (QDT) withhis invention of The Canary System. Based on similarprinciples, it provides information on the condition of thetooth structure. Unlike fluorescence-based devices thatreveal the presence of bacteria close to the tooth’s surface,The Canary System identifies abnormalities in the crystalstructure of the tooth up to a depth of 5mm. It can alsopinpoint cracks that are causing pain and sensitivity. Noother caries detection device on the market has these ca-pabilities [4]. With Prof. Mandelis and his team’s ongoing

consistent effort in this field, we should confidently expectthe realization of Baymax in our generation.

References[1] “Radiology Safety,” Available: http://www.

radiologyinfo.org/en/safety/?pg=sfty_xray. [Ac-cessed: 2015 Feb 23].

[2] “Bioacoustophotonic Depth-Selective Imag-ing of Turbid Media and Tissues,” Available:http://www.wseas.us/e-library/conferences/2006elounda1/papers/537-284.pdf. [Accessed: 2015Feb 23].

[3] “The Killam Prize Winner Press Release,” Avail-able: http://www.canadacouncil.ca/council/prizes/find-a-prize/prizes/killam-prizes. [Accessed: 2015Feb 23].

[4] “Prof. Mandelis, Co-Funder of the QDT Wins The Kil-lam Prize,” Available: http://www.thecanarysystem.com. [Accessed: 2015 Feb 23].

galbraith society undergraduate engineering journalalice.ye/... · galbraith society club meeting:...

Documents

capital area genealogical society 3/20/2012 cemetery ......

mathematical and astronomical tables, galbraith

from galbraith to krugman and back galbraith, krugman and...

barry galbraith - fingerboard workbook.pdf

ﬁcult airway society 2015 guidelines for management of...

barry galbraith chord melody

james k. galbraith - utip

robert galbraith: silkeormen

barry galbraith guitar comping

barry galbraith - guitar comping - book

james kenneth galbraith galbraith@mail.utexas.edu...

the “affluent society” of john kenneth galbraith

the affluent society. american abundance john kenneth...

james kenneth galbraith...

barry galbraith - guitar comping

james k. galbraith

tree safety ckd galbraith

barry galbraith - comping.pdf

gwutjoe galbraith

galbraith guitarist article