TABLE OF CONTENTS

1. introduction and theory

During the last 50 years, dozens if not hundreds of aquifers across the continental US have been contaminated with various halogenated and oxy-halogenated hydrocarbons, largely through military and industrial operations. Hence, a significant amount of work has been done on the development of technologies for the remediation of such groundwater. The efficacy of bimetallic catalysis for dehalogenation and reduction of such compounds is well established. The mechanism effecting catalysis (lowering of the reaction E_A) often involves electron and hydrogen transfer through the ZVI to the CHC being reduced. It may also be enhanced both by the resulting potential across the interface of the two metal species arising from the difference in the standard electrode potentials of the two metals and by other forces which tend to strain the carbon-halogen covalent bonds. In the case of Iron / Palladium catalyst particles, the standard electrode potential (hereafter SHE) of palladium is +0.987 V SHE and that of Iron (Fe) is –0.440V SHE. This means that when the two metals are brought into intimate contact, the resulting bimetallic voltage at their interface is approximately Del V= 1.427 V. The combined effect of these forces is often to break the carbon-halogen bond(s) in question, cleaving the molecule and releasing the halogen as a radical or ion and causing the olefin to hydrogenate or hydroxylate at that site or form a higher-level carbon - carbon bond with a neighboring vinyl group. The halogen radicals and ions then tend to readily combine with one another, hydrogen ions, or metallic species in the water to form diatomic halogen species, acids, and salts respectively, all of which have very low specific Gibbs free energies and are thus extremely stable. In general, the resulting species are more environmentally benign and remediation is effected. In order for the reaction to take place, however, the hydrocarbon pollutant species must come into intimate contact with the bimetallic interface. When the pollutant exists in an aqueous solution at the ppm or ppb level, however, the transport of the molecule to the bimetallic interface of a standard-sized catalytic particle becomes (by far) the rate-limiting step, and previous attempts to remediate polluted water have not been practicable because the product “P” of

P = (AREA OF BIMETALLIC INTERFACE AVAILABLE x REACTOR RESIDENCE TIME x INLET POLLUTANT CONCENTRATION)

has been far to small , and most of the pollutant molecules passing through a typical packed reactor never actually adsorb onto the interface that they may be broken down. The only way to achieve satisfactory overall conversion is to greatly increase the product “P”. Since decreasing the pollutant concentration is the object of remediation and increasing residence time means making larger (high capital-cost) systems, the only way to cost-effectively increase P is to increase the bimetallic area available in a given amount of reactor volume by making and using (i.e. immobilizing within a reactor) very small bimetallic catalyst particles. So when nano-scale bimetallic nanocatalysts with a tight size distribution were successfully made by Golder and associates, the authors set out to find a way to immobilize and expose a very large number of these bimetallic nanoparticles (hereafter BNP’s) in a small volume, that each was available to contact over nearly all of it’s surface area AND fixed in place such that it would not become entrained in the flow through the reactor. Once a suitable method of immobilization and fixation was developed and tested, the feasibility of this ‘pump-treat-and-use’ scheme was greatly enhanced, and the preliminary tests herein were warranted. Since the method of immobilization was not conducive to accurate modeling, an empirical approach to the tests was adopted: A fully-blocked (non-orthogonal) matrix of test runs was first proposed to explore the efficacy of each of two reactors in remediating small batches of artificially polluted water under varying conditions. The parameters varied included: two values of flowrate to study the efficacy of the system at different residence times and two different inlet contaminant concentrations. Varying these data would have varied the above-mentioned product “P” to provide insight into the true design constraints determining the economic feasibility of differing strategies for site remediation via this pump-and-treat method. The protocol, however, called for some 96 runs to be made for each reactor, totaling 192 runs and requiring 240 separate sample analyses to complete the study. At an approximate average cost of ~$125 per analysis, the projected cost of obtaining response data from such a test (~$30,000.) was deemed prohibitive, and a much less comprehensive experiment was instead carried out.

2. Apparatus

Referring to figure 1, the apparatus included a covered feed container into which a submersible pump was lowered. The pump pumped the fluid from the feed container through the reactor , through a variable area flow meter, and into the open discharge container. The discharge and feed containers were interchangeable to facilitate additional ‘passes’ through the reactor (second- and third-stage treatments) without exposing the water to the atmosphere unnecessarily, effectively allowing us to study , once again, different values of our product “P”.

2.1 Figures

Figure 1 – Schematic of the equipment used in the pilot process

3. Procedure

1) Protocol

a) PM reactor was challenged with two pollutants. Then the EM reactor was challenged with three pollutants, making 5 runs in all, as depicted in the following experiment Matrix:

Contaminant Species	Reactor Type	Pass #	Flowrate, GPM

Trichloroethylene	PM	1	1
		2	1
		3	1
Ammonium Perchlorate	PM	1	1
		2	1
		3	1
Trichloroethylene	EM	1	0.28
		2	0.26
		3	0.25
Ammonium Perchlorate	EM	1	0.28
		2	0.25
		3	0.24
N-Nitrosodimethylamine	EM	1	0.22
		2	0.25
		3	0.25

2) Method

a) The aqueous BNP solution was diluted by 50 % due to H2O evaporation which had taken place after purchase and before testing: 250 ml filtered H2O was added to the 250 ml NP solution remaining in the original bottle. The bottle was then vigorously shaken to homogenize the BNP’s

b) The PM reactor was prepared for service and connected into reaction system

i) The HENCI reactor was prepared

ii) The reactor was sealed off by closing it’s ISO valves

iii) 3l water added to sealed reactor

iv) 100 ml NP solution was added to sealed reactor

c) The recycle tube was then plumbed in (bypassing flow meter , and leading back into the dummy feed bucket filled with flush-water)

d) ISO valves were opened

e) Pump was turned on and flush water recycled through reactor for 5 minutes to allow all NPs to become bound in media

f) Pump turned off and ISO valves closed

g) Flush water allowed to drain from system. About 1 l flush water remains in reactor and tubing.

h) Switched feed bucket to that with the pollutant solution

i) Turned on pump and recycled pollutant solution directly back into feed bucket for 30 to 60 sec

j) Took sample of the feed bucket solution for analysis

k) Conduct First Pass test

i) Close recycle valve and open ISO valves to begin flow through the rector

ii) Adjust flow meter valve to achieve desired flowrate

iii) Allow reactor discharge to collect win discharge bucket until feed bucket is empty

iv) Sample from discharge bucket and label sample appropriately ( for whichever pass)

v) Discard feed bucket

vi) Carefully bring discharge bucket over to feed-side of the apparatus, insert the submersible pump, and close the lid. This bucket is now the feed bucket

vii) Install a new bucket under the discharge tube of the process. This bucket is now the discharge bucket

l) Conduct Second Pass

i) Repeat steps k)I through k) VII above

m) Conduct Third Pass

i) Repeat steps k)I through k) VII above

n) Clean all equipment and repeat steps c) through m) above for the next pollutant ( see matrix)

o) The EM reactor was prepared for service and connected into reaction system

i) The HENCI reactor was prepared

ii) The reactor was sealed off by closing it’s ISO valves

iii) ~1l water added to sealed reactor

iv) Reactor BNP retention system was enabled

v) 30 ml BNP solution was added to sealed reactor

p) Steps c) through n) were repeated for the EM reactor for it’s first pollutant challenge (see test matrix)

q) Steps c) through n) were repeated for the EM reactor for it’s second pollutant challenge (see test matrix)

r) Steps c) through n) were repeated for the EM reactor for it’s third pollutant challenge (see test matrix)

s) Equipment was cleaned, disassembled and packed up

4. Results

The analyses of the 20 samples along with a few calculated values are summarized below. Unfortunately, for reasons beyond the scope of this paper, only the third quartet of data (TCE on the EM reactor - highlighted) is valid response data. Hence, all discussion of the results will be limited to this sole valid dataset.

Input Variables and Raw Data							Calculated Values
Contaminant Species	Reactor Type	Flowrate, GPM	Calc'd Inlet Conc, mg/l	Measured Inlet Conc, mg/l	Pass #	Measured DischargeConc, mg/l	Reaction Efficiency (% conversion)	Overall (3-pass) Percent Conversion

Trichloroethylene	PM	1	148.000	Data Not Valid	1st	Data Not Valid	N/A
		1		Data Not Valid	2nd	Data Not Valid	N/A
		1		Data Not Valid	3rd	Data Not Valid	N/A	N/A
Ammonium Perchlorate	PM	1	~120	Data Not Valid	1st	Data Not Valid	N/A
		1		Data Not Valid	2nd	Data Not Valid	N/A
		1		Data Not Valid	3rd	Data Not Valid	N/A	N/A

*Trichloroethylene*	EM	*0.28*	*148.000*	*14.200*	*1st*	*10.300*	*27.46*
		*0.26*			*2nd*	*6.970*	*32.33*
		*0.25*			*3rd*	*5.080*	*27.12*	*64.23*
Ammonium Perchlorate	EM	0.28	~100	Data Not Valid	1st	Data Not Valid	N/A
		0.25		Data Not Valid	2nd	Data Not Valid	N/A	N/A
		0.24			3rd	89.000	-1.14	5.32
N-Nitrosodimethylamine	EM	0.22	100.004	Data Not Valid	1st	Data Not Valid	N/A
		0.25		Data Not Valid	2nd	Data Not Valid	N/A
		0.25		Data Not Valid	3rd	Data Not Valid	N/A	N/A

5. discussion

1) Data Trend Analysis

a) TCE on the PM Reactor – no valid response data

b) Ammonium Perchlorate on the PM Reactor – no valid response data

c) TCE on the PM reactor:

i)
Overview: Evidently, in the TCE run on the PM reactor, each pass through the reactor decreased the concentration of Trichloroethylene by approximately 30% (see data table). This means that the combined effect of evaporation and catalytic breakdown of TCE during each test run was to decrease the concentration of TCE by ~30 % per pass.

ii) Reaction and Mechanism: Since the reaction of interest is the breakdown of TCE, we shall look at the reaction kinetics and see if this data can corroborate a valid kinetic model. First let’s look at the reaction and it’s mechanism(s).

(1) For the purposes of this report, we’ll assume that the pH of the water is between 5 and 9 so that no Lewis-acid/base reactions occur with a high enough rate to compete with the cleavage reaction itself. It is the author’s understanding that this assumption is most often valid within the groundwater remediation industry.

(2) In this regime, when TCE is exposed to the bimetallic interface and chlorine is removed from the olefin, it may well receive an extra electron from the (cathodic) palladium or zero-valent iron during, or to facilitate, the cleavage reaction. In that case, the resulting chlorine anion (Cl^-) may be more likely to take place in an ionic combination: It may recombine with (1) a metal cation in solution to create a salt (a fairly benign fate for Cl in most treatment applications), or (2), a hydrogen cation to create the Lewis acid HCl, and the overall effect of the extra electrons from the palladium may be to lower the pH of the effluent, if it is indeed high at the reactor inlet (another positive outcome or ‘fate’ for the Cl). If it does not, however, receive an extra electron upon cleavage and it ‘comes off’ as a radical (Cl^.) , then it is more likely to recombine into Cl2 gas. In either case, however, note that the olefin itself has not had to supply the extra electron, and therefore is likely to remain an alkene: that is , the double bond will likely not be broken to a single (alkane) bond. However, if the Palladium does NOT act as a true cathode, then it is quite possible that the double bond will be reduced to a single C-C bond and an alkane (ethane) can be formed. This means that there are now six free bond-sites to be filled on each ethane molecule created. If this DOES happen, and if the pH of the solution is high enough, it is quite conceivable that one or more of the free sites will be replaced with hydroxyl ions instead of hydrogen ions, resulting in an alcohol or poly-ol, some of which, like ethanol, may be more benign than ethane. Other (higher-order) polyols, however, may be less benign than ethane.

(3) In either case, the bond-order of the resulting aliphatic hydrocarbon is much less toxicologically important than the degree of dehalogenation: it is the conversion to a completely dehalogenated species which is most important: the partial dehalogenation of TCE to dichloro- or the (less likely) chloro-ethene (hereafter DCE and MCE) is of little good, as these compounds are also carcinogenic and poisonous. Hence, one of the biggest weaknesses of this study is that we do not assay for DCE or MCE: hence from our results, we can only ascertain how much TCE is gone – not what we have replaced it with. In other words, we can only calculate percent breakdown, not percent conversion.

(4) With a bit of up-stream (automated) pretreatment and more research, however, the potential to convert TCE very completely to very benign breakdown byproducts and to verify the conversion is very real

(5) In any case, for our purposes here, let’s look at the portion of the cleavage reactions that DO go to completion, and are re-substituted by only hydrogen, resulting in the completely unsaturated byproduct ethylene, leaving study of conversion optimization for a future round of tests.

(6)

The overall cleavage reaction can be written thusly: 2(C₂Cl₃H) + 6H⁺ ® 2(C₂H₄) + 3Cl₂ .

Now, the reaction mechanism may well look something like this:

(1) C₂Cl₃H ® C2H +3Cl

(2) C2H + 3H+ ® C2H4

And therein lies the rub: which of the two is the slowest (rate-limiting) step? If step (1) is rate-limiting, then the reaction kinetics can be estimated as first-order, meaning that the rate of conversion of TCE into ethylene is simply proportional to the concentration of TCE in the bulk of the reactor, and hence, at steady-state operation, proportional to the concentration of TCE in the feed stream. If however, step (2) is rate-limiting, then the conversion rate will be proportional to the inverse log of the pH of the feed stream. Since we’ve no control over the initial feed stream of polluted groundwater, and furthermore, whereas we could adjust the pH of our reactor inlet stream quite easily with acid or base injections, then for the purposes of this discussion, let us explore what kind of data correlation we get assuming that pH of the inlet stream is constant at some value which provides enough hydrogen ions in solution so as to make step 1 above the rate-limiting step

iii) Kinetics: In this case, then, we have the dehalogenation itself as the rate-limiting step. Now, since the cis-and trans- Cls has differing bond-energies arising from the stereochemistry of the molecule, they will not all be cleaved with the same ‘ease’. Hence, the actual mechanism may well include DCE and MCE intermediaries, as the TCE loses one, then two, then three of it’s chlorine ‘ligands’ sequentially. Assuming, however, that the cleavage reaction does go to completion at each reaction site ( the interface of Pd and Fe on the nanoparticle surface) and all three chlorine ligands are removed before the hydrocarbon leaves (desorbs or ‘ vibrates off’) the BNP surface and back into solution, then we can develop a model for the (first-order) kinetics of the reaction: The rate equations can be written thusly:

-d[TCE]/dt = k x [TCE] which , upon rearrangement yields

d[TCE]/[TCE] = -kdt which upon integration yields

ln[TCE] – ln[TCE]₀ = -k(t-t₀) = -kt which, upon rearrangement yields

ln[TCE] = ln[TCE]₀ – kt which, upon taking the inverse natural log of each side, becomes : [TCE] = [TCE]₀ x e^-kt

Note that this is the typical Ahhrenius form of first order decay kinetics: The rate of conversion decreases as the amount of TCE available for conversion decreases. Now, if we try substituting values of, say 18 ppm for [TCE]₀ , and 0.3 for k, and then graph this equation, we get the familiar first-order ‘Ahhrenius’ curve with the following values:

Note how similar in shape (and in values!) this curve looks to the data taken at the MNWD for the EM reactor breaking down TCE. The main difference is that this model would be best applied to a batch reactor wherein the TCE solution was simply exposed to the catalyst for an extended period of time (called ‘batch exposure time’) without flowing, and then the perfect ‘average’ sample was magically taken out of the solution every so often. This situation is effectively mimicked, however, by our flow-through scheme (called a continuous reactor as opposed to a batch reactor) in which the samples were taken after each ‘run’.

iv) Process Engineering Analysis and Scale-up: In fact, it is important to understand the continuous-reactor’s counterpart to ‘batch exposure time’: the duration of exposure for each molecule of water or TCE as it flowed through the reactor. This is called the reactor ‘Residence Time’, and is only completely valid number when the reactor is operated at fully developed (turbulent or ‘plug’) flow , which only occurs when the reactor’s Reynolds number (pvd/u) is over 2300. In our reactor, however, we can relax this requirement because physical conditions in the reactor render the flow effectively turbulent at low superficial values of the Reynolds number, though without increasing the reactor’s overall pressure drop enough to impact commercial applicability. Further testing can determine the optimal values of flowrate and reactor length, not to mention a host of other parameters beyond the scope here. For our purposes here then, we’ll use the median value of residence time to represent the true residence time of the reactor. The residence time is then the volume of the reactor divided by the flowrate:

Residence Time (Tr) = Reactor Volume / Flowrate. Since the valid data set was obtained on the EM reactor, we’ll calculate the Tr for the EM reactor. The EM reactor volume is 12.54 in³ = 0.0543 US gal, but our best estimate of the portion of the reactor volume occupied by the packing and NPs is about 6%, so the actual reactor volume is 0.0543 x 0.94 = 0.05103 gal. Now, from the data table above, we know that the flowrate through the EM reactor averaged approximately 0.28 GPM, so Tr = 0.1483 min, or about 8.9 seconds. Hence, we achieved approximately 30% removal of TCE (evaporation as well as reaction) in about 9 seconds of total exposure at an inlet concentration of about 14 ppm. Furthermore, the percent conversion at the lower inlet concentrations in the data (the last run achieves ~30% removal at about 5 ppm / 14 ppm or 36% of the inlet concentration) stays about the same, suggesting that the conversion of TCE to ethane and Cl2 gas is not mass-transport-limited at these inlet concentrations. If this also holds true at concentrations of about 1/3^rd of what we now see, we’ll know that we can take 1.4 ppm down to less than 14 ppb in `10-12 reactor lengths or about 15 - 20 ft of total reactor length, which can easily be folded or modularly configured into a system small enough to fit on the back of a small truck. Of course, scaling up the nanoparticle retention per unit reactor volume and the diameter of the reactor are also feasible and would result in significant increases in conversion. For instance, if we assume that doubling the nanoparticle loading (surface area available for reaction / vol. of the reactor) would increase the rate of destruction by even 80% (a safe bet), and we also increase the diameter of the reactor to, say, 1 ft, then , nominally, due to the increased conversion and throughput, it turns out that for this particular scale-up, an 8-ft long , 1 ft diameter reactor would be able to remediate ( to < 1.0 ppb) approximately 35 GPM of 1.4 ppm TCE inlet water. This now represents a sensible and economically feasible pump-and-treat solution which is effective at the pollutant levels found in groundwater, and can be mobilized on a small truck.

v) Other Considerations: These results seem commensurate with previous research on the in-situ remediation of TCE with Fe / Pd nanoparticles (see Zhang, Wei-xian, Journal of Nanoparticle Research, 5:323-332, 2003) which attests to the ability of these nanoparticles to remediate TCE (and similarly) polluted groundwater to <10 ppb TCE within 8 hrs of batch exposure at a nanoparticle loading concentration of 6.25g BNP’s per liter of polluted water. The extended time period is necessary to allow each TCE molecule in the quiescent body of water (whether it be an aquifer or a beaker) to come into contact (through random Brownian motion) with the nanoparticle bimetallic interface. The configuration of the continuous reactors used herein provide a ‘boost’ to that Brownian- (and concentration gradient-) driven contact by greatly increasing the coefficient of mass transport for the reaction system in the same way that a packed column increases overall mass transport over a non-stirred batch reactor.

b) Ammonium Perchlorate on the EM Reactor – no valid response data

c) Nitrosodimethylamine on the EM Reactor – no valid response data.

6. Conclusions

i) Though it is becoming widely known that Bi- and even mono-metallic zero-valent Fe-containing nanoparticles hold much promise for groundwater remediation, with the new method of immobilization used herein, new life may be afforded to pump-and treat strategies for groundwater remediation.

ii) Specialized bimetallic nanoparticles also have a great potential in general catalysis in the chemical process industries. The proprietary nanoparticle immobilization techniques used in this study will be of great value to an assortment of high-tech and Ultra-High-Purity chemical process-based industries.

iii) In order for a continuous reactor approach to the exploitation of BNP catalytic efficacy to be truly advantageous, the continuous reactor configuration must be correctly designed to affect intimate contact of the fluid to the BNP’s with each pass through the reactor body. In this way only, the continuous reactor has the effect of greatly increasing the effective loading of BNP’s per unit volume of water without actually using up the BNP’s, and the overall rate of chemical conversion is greatly enhanced

7. Acknowledgements

Thanks goes out to Malcolm Jones,P.E.

8. References

Zhang, Wei-xian, Journal of Nanoparticle Research, 5:323-332, 2003