An inorganic aqueous solution, known as IAS, has shown its compatibility with aluminum phase-change heat transfer devices. When using IAS in aluminum devices, aluminum prefers to react with the two oxidizers, permanganate and chromate, rather than water to generate a thin and compact layer of aluminum oxide, which protects the aluminum surface and prevents further reactions. In addition, an electrochemical theory of aluminum passivation is introduced, and the existence of an electrochemical cycle is demonstrated by an aluminum thermosiphon test. The electrochemistry cycle, built up by liquid back flow and tube wall, allows the oxidizers to passivate the aluminum surface inside the device without being directly in contact with it. However, failure was detected while using IAS in thermosiphons with air natural convection cooling. The importance of a continuous liquid back flow to aluminum passivation in phase-change heat transfer devices is pointed out, and a vertical thermosiphon test with natural convection cooling is used to demonstrate that a discontinuous liquid back flow is the main reason of the failures.

Introduction

Nowadays, with the rapid development of science and technology, the amount of heat produced by electronic components during operation has been increasing rapidly. Phase-change heat transfer devices, such as heat pipes, are widely used for electronics cooling because of the high effective thermal conductance. A phase-change heat transfer device usually has a vacuumed internal space which is charged with a small quantity of working fluid. Because of the low internal pressure, the working fluid evaporates at much lower temperature compared to what it would be in the open air. When heat is added to the evaporator, the working fluid evaporates and takes away the heat. The vapor flows through the internal space because of the pressure difference resulting from a lower temperature at the cooled section and condenses and releases the heat to the condenser. The condensed liquid is returned from the condenser to the evaporator by gravity or capillarity.

For ground-based phase-change heat transfer devices, copper and water are the most commonly used casing material and working fluid. Water is the most common liquid, because it has a large latent heat of evaporation. Copper is chosen because of its high thermal conductivity and excellent compatibility with water. However, this combination is not perfect anymore because of the weight limit in some industries. Aluminum is widely used due to its three times less density than copper [1], but it is not compatible with water. Aluminum is a high chemically active material, and it can reduce water and generate hydrogen gas, a noncondensable gas (NCG), which can fail the phase-change heat transfer device in a few minutes. As a result, fluids such as ammonia or methanol are used as the working fluid in aluminum phase-change heat transfer devices to avoid generating NCG. However, because of their low latent heats of evaporation, the maximum load of such devices is limited.

Chromate is found to be a corrosion inhibitor of aluminum alloys by Kendig and Buchheit [2], and Rocco et al. [3] compared several methods of chromate coatings on 55% Al–Zn coated steel, and he concluded that chromate coatings increased corrosion resistance and allowed water to be in contact with an aluminum surface and actively resisted oxidation. However, most fail when there is heat transfer present.

A fluid known as inorganic aqueous solution (IAS) has shown its compatibility with aluminum phase-change heat transfer devices, as demonstrated by Reilly et al. [4]. IAS fluid is an aqueous solution with the six inorganic chemical constituents, as shown in Table 1. One method to produce IAS is dissolving solid potassium permanganate, potassium dichromate, chromium trioxide, calcium hydroxide, strontium hydroxide, and sodium hydroxide into water. Fresh IAS fluid has an initial pH of 6.22. Aluminum passivation is the only concern of this paper, so sodium, potassium, calcium, and strontium ions, which are designed to improve surface wettability [5], will not be discussed in this paper.

Aluminum Passivation

In general, aqueous solutions are not able to be used as the working fluid in aluminum phase-change heat transfer devices. Aluminum metal is highly chemically active, and it is easily oxidized by water to become aluminum (III) compounds. In addition, the reaction generates hydrogen gas, a NCG 
2Al(s)+6H2O(l)Al2O3(s)+3H2(g)
(1)

which will be compressed and accumulated in the condenser of the device forming a NCG block. Over time, the gas block grows, reduces the effective condensing region, and eventually fails the heat transfer device, see Fig. 1.

However, aluminum has been shown to be passivated by several different methods in aluminum corrosion prevention researches. No matter which method is chosen, the final goal is generating a thin and compact layer of aluminum oxide to protect the surface and prevent further reactions. One of the methods to generate such a layer is locating the surface in a strong oxidation environment, and the oxidizers in the IAS fluid are strong enough to make it happen.

Figure 2 [6] is an E-pH diagram, also called Pourbaix diagram, of aluminum at 25 °C in an aqueous solution. Lines (a) and (b) are the boundaries corresponding to the stability of water. Above line (b), water will be oxidized, and oxygen gas will be generated. Below line (a), water will be reduced, and hydrogen gas will be generated. When aluminum is in contact with pure water, water will be reduced to generate hydrogen gas because the region of aluminum immunity is much lower than line (a).

Moreover, when aluminum is in contact with an aqueous solution, if the solution is strong acidic or basic, aluminum (III) ion or aluminate ion will be generated separately, as seen in Fig. 2. However, if the aqueous solution is neutral, aluminum (III) will be generated as aluminum oxide, a thin and compact layer of coating, which will protect the aluminum surface and prevent further reactions.

Aluminum oxide can protect the aluminum surface and prevent further reaction, but it is not generated as a single whole piece. For an aluminum surface precoated by aluminum oxide, ions can still be transported through the gap of two pieces of aluminum oxide and react with the aluminum surface, but the overall reaction rate is tremendously reduced. As a result, such an aluminum surface can be used directly in contact with water in open air and has little corrosion. However, if the pretreated aluminum surface is used in a phase-change heat transfer device with water as the working fluid, a small amount of hydrogen gas will still be generated slowly, and it is irreversible. Because of the low pressure in phase-change heat transfer devices, the volume of the small amount of hydrogen gas is not negligible, and it can fail the device in hours or days. This is why a pretreated aluminum phase-change heat transfer device is still not compatible with water. Nevertheless, if the aqueous solution is oxidative, above line (a), because of a soluble strong oxidizer, aluminum prefers to react with the oxidizer rather than water, which will prevent the generation of hydrogen gas.

Therefore, in order to make aqueous solutions compatible with aluminum phase-change heat transfer devices, an approach can be given that the potential and pH number of the solution should locate in the region of passivation but bounded by line (a) and line (b) in Fig. 2. There is no hydrogen gas or oxygen gas, both of which are NCGs, generated, and the corrosion of aluminum is greatly reduced because of the protection of aluminum oxide. In addition, products of the reaction should have no gases.

However, the lines move with temperature. For aluminum phase-change heat transfer devices, the region in Fig. 2 that guarantees passivation and NCG prevention, while both being stored and operating, will be smaller. Stubblebine et al. [7] studied how the passivation region, for aluminum phase-change heat transfer devices, changes with temperature.

Inorganic aqueous solution fluid is such an aqueous solution with the potential and pH number in the passivation region during the lifetime of the heat transfer devices, and products generated by reactions between IAS and aluminum are water and solid compounds. The electrode potential and pH number of a fresh IAS fluid are 0.910V (SCE) and 6.22 separately, see point A in Fig. 2 [6], and the ones of a yellow fluid, the IAS fluid collected from used aluminum heat transfer devices, are 0.233–0.245V (SCE) and 6.65–7.03 separately, see point B in Fig. 2 [6]. If IAS is used as the working fluid in aluminum devices, because of the strong oxidizers in IAS fluid, aluminum prefers to react with the oxidizers first instead of hydrogen ions. The two reactions, Eqs. (2) and (3), form a thin and compact layer of aluminum oxide that protects the surface and prevents further reactions 
2Al(s)+2MnO4(aq)+2H+(aq)Al2O3(s)+2MnO2(s)+H2O(l)
(2)
 
2Al(s)+2CrO42(aq)+4H+(aq)Al2O3(s)+Cr2O3(s)+2H2O(l)
(3)

For aluminum, both permanganate and chromate are oxidizers. Permanganate is stronger and reacts faster, so it is useful in the initial passivation of the aluminum surface. Usually, it is reacted out several hours after the device is charged. However, the slower reacting but larger amount of chromate is available for the oxidation to heal the aluminum oxide layer and maintain the passivation over the lifetime of the device.

Passivation in Phase-Change Devices

Because of the strong oxidizers, IAS can passivate aluminum surfaces with no NCG generated while aluminum surfaces are directly submersed in IAS. However, the chemicals in IAS are inorganic, which have much higher boiling points than water. As a result, the vapor of IAS is pure water vapor. In an operating phase-change heat transfer device, there is only pure water condensed in the condenser. In addition, because advection is dominant in phase-change heat transfer devices, all the chemicals are pushed to the evaporator [5]. As a result, the aluminum surfaces in the condensing and adiabatic regions are still exposed to pure water. IAS is found to be compatible with aluminum phase-change heat transfer devices [4]. What makes the aluminum surface in condensing and adiabatic regions not react with the pure water? How does IAS passivate the aluminum surface and prevent the generation of hydrogen gas in the condensing and adiabatic regions?

Electrochemical Cycle in Aluminum Passivation.

The existence of an electrochemical cycle in an operating aluminum phase-change heat transfer device can be used to explain how IAS passivates the aluminum surface and prevents the generation of hydrogen gas in the condensing and adiabatic regions, see Fig. 3.

In the evaporating region, a thin and compact layer of aluminum oxide is generated to protect the surface. Aluminum oxide is an electronic insulator, but it is not generated as a single whole piece. Electrons and ions can still be transported through the gap of two pieces of aluminum oxide, and it is also where all the redox reactions take place.

As a result, once a bare aluminum surface is exposed in the condensing or adiabatic region, one aluminum atom loses three electrons and then combines with three hydroxide ions to generate aluminum oxide, see reaction (4), which passivates the bare aluminum surface. The electrons are transported by the aluminum body of the device, go through the gap of the aluminum oxide, and then react with the oxidizers in the reservoir. On the liquid side, the residual hydrogen ions are transported back to the evaporating region by the liquid back flow to contribute to the reactions between the oxidizers and the electrons, as shown in reactions (5) and (6). As a result, to accomplish passivation of the whole device, a continuous liquid back flow is required to cover the whole internal surface 
2Al(s)+3H2O(l)6eAl2O3(s)+6H+(aq)
(4)
 
MnO4(aq)+4H+(aq)+3eMnO2(s)+2H2O(l)
(5)
 
2CrO42(aq)+10H+(aq)+6eCr2O3(s)+5H2O(l)
(6)

The electrochemical cycle enables an aluminum surface to be passivated by IAS while not being in contact with it.

Aluminum Thermosiphon Test.

An aluminum thermosiphon test was performed to demonstrate the existence of the electrochemical cycle. Figure 4 is a schematic of the aluminum thermosiphon test. It is composed of a cartridge heater block, an ice water condensing reservoir, and an aluminum 6061 tube with one end welded with an aluminum joint end cap and the other end closed by a Swagelok bellows-sealed valve. The OD of the aluminum tube is 0.375 in, and the ID is 0.311 in. Before charging, each thermosiphon is pretreated by 2 mol/L hydrogen chloride solution to remove aluminum oxide and then rinsed by di-water.

Figure 5 shows the thermocouple locations in the aluminum thermosiphon experiment. There is a total of ten type T thermocouples used to monitor the temperatures: two in the evaporating region, two in the condensing region, and six in the adiabatic region.

Figure 6 shows the performance of water and IAS in aluminum thermosiphons. Average temperature of thermocouples in each section is used. For each run, the test was initiated with an input power of 20 W. Steady-state was reached in 15 mins. The input power was then increased by 10 W every 10 mins. After 85 mins, the input power of the water filled tube was maintained at 90 W. However, the input power of the IAS tube continued being increased every 10 mins until a value of 160 W was reached.

The result indicates that NCG was obviously generated in the water-filled thermosiphon when the evaporating temperature was greater than 60 °C, and finally failed the test. Even after the input power stopped increasing, the temperature difference between the evaporating and the condensing regions kept becoming larger. However, for IAS fluid, no NCG was generated. The test was stopped before the evaporating temperature reached 120 °C because of a safety consideration of the welded cap. No dry out or other thermal limitation was noted. The test clearly demonstrates that IAS is compatible with aluminum heat transfer devices, and that no NCG is generated. Moreover, the adiabatic temperature was as high as 90 °C, which is much higher than 60 °C. It means that hot water at 90 °C was directly in contact with the aluminum surface in the adiabatic region, but there was no NCG generated. Therefore, the existence of the electrochemical cycle has been demonstrated.

A copper thermosiphon, with the same geometry and charged with water, was tested under the same conditions as the aluminum/IAS one. Figure 7 shows the comparison of the heat transfer performance between them. The temperature difference (ΔT) of the evaporator and the condenser was measured and compared.

It shows that the heat transfer performance of an aluminum/IAS thermosiphon is close to the performance of a copper/water thermosiphon. Therefore, aluminum/IAS thermosiphons have the potential to replace copper/water thermosiphons in the market.

Importance of a Continuous Liquid Back Flow

For a thermosiphon, if the condensation rate in the condenser is insufficient, the condensed liquid exists as droplets instead of a continuous liquid film, so a discontinuous liquid back flow appears, see Fig. 8 [8]. In this case, residual hydrogen ions will be accumulated in the disconnected part, and an electronic force will be generated to prevent electrons from transporting away. If the concentration of the residual hydrogen ions surpasses a limit, electrons will choose to react with the hydrogen ions and generate hydrogen gas, an NCG which eventually fails the heat transfer device. With the input power increasing, the disconnected region will shrink and eventually disappear.

As a result, how a thermosiphon is cooled is important to the existence of a discontinuous liquid back flow. Figure 9 shows three thermosiphons with the same geometries but heated by different input powers or cooled by different methods. Only the condensing and adiabatic regions are shown. Tube (1) is cooled by a cooling block, and the rest part of the tube is adiabatic. Tubes (2) and (3) are cooled by natural convection, and there is no adiabatic region. Tubes (1) and (2) are heated with an input power of Q, but tube (3) is heated with an input power of 2Q.

It can be seen that for a given Q, tube (2) has the largest discontinuous liquid flow region, in which liquid is disconnected with the electrochemical cycle and hydrogen gas will be generated. In addition, with Q increasing, all the disconnected regions shrink, and tube (1) will be the first one obtaining a continuous liquid back flow throughout the whole tube.

Aluminum Thermosiphon Test With a Natural Air Convention Cooling.

An aluminum thermosiphon experiment with a natural air convention cooling is setup to demonstrate the continuous liquid back flow requirement. Figure 10 is a schematic of the experiment setup. An aluminum 6061 tube, with 0.375 in OD and 0.311-in ID, is selected as the body of the thermosiphon. One end is welded with a joint end cap, and the other end is closed by a Swagelok stainless steel bellows-sealed valve. A copper heater block with two cartridge heaters is used as the heat source, and the thermosiphon is cooled by the natural air convection. Four tests were done with tube lengths of 1 ft, 2 ft, 3 ft, and 6 ft. All the thermosiphons are charged with 1.4 ml IAS fluid, which covers about 60% of the evaporating region.

Figure 11 shows the thermocouple locations of the aluminum thermosiphon test. Two thermocouples are located in the evaporating region. Six thermocouples are located in the first 9 in from the bellows-sealed valve, and additional thermocouples are added every 9 in. All the thermocouples are numbered from the top to the bottom. Because failure or not is the only concern in this test, the heat transfer performance will not be quantitatively discussed.

Based on the continuous liquid back flow theory, for a fixed input power, with tube length increasing, discontinuous liquid back flow appears after a critical length. Moreover, with the tube length further increasing, the length of the discontinuous liquid back flow region increases. Therefore, if a 1-ft thermosiphon can hold the aluminum passivation, with the tube length increasing, there must be a critical length that thermosiphons with a smaller length than it will hold the passivation, but the ones with a larger length than it will fail. In addition, the longer tube length is than the critical one, the faster NCG is generated.

Test Results and Discussion.

A 6-ft thermosiphon was first tested. Fifety watts were added for the first 20 mins, and then increased to 100 W. Figure 12 shows the test results of the 6-ft thermosiphon. After 16 mins, T1 started to decrease. It was a sign of the NCG appearance. After 20 mins, a total of five thermocouples were covered by NCG, and then the input power was increased to 100 W, and T4 and T5 went back to normal, which is another sign of NCG generation, because a higher vapor pressure, caused by the higher temperature, compressed the NCG. With the tube length decreasing, the area of the condensing region decreases, so the surface temperature increases. In order to avoid the internal pressure being too high, the input power was limited at 50 W in the later tests. For the set charge amount and input power (50 W), 6 ft was found to be larger than the critical length.

Three-foot and 2-ft long thermosiphons were tested afterward, and the input power was 50 W. As shown in Fig. 13, the 3-ft thermosiphon is definitely a failing case, so it is still larger than the critical length. In Fig. 14, T1 is the only one that is different with the other temperatures in the condensing region, but it is still close. As a result, 2 ft is still larger than the critical length, but the critical length is very close to 2 ft.

Figure 15 shows the test result of a 1-ft aluminum thermosiphon. The temperature difference of the seven thermocouples on the condensing region is within 1 deg, which is within the accuracy of the thermocouple. Therefore, it is assumed that there is little to no NCG generated, and 1 ft is likely smaller than the critical length.

The charge amount was fixed at 1.4 ml for all the tests. With the tube length, and therefore, the internal surface area, increasing, it definitely requires more oxidizers to finish the passivation process. Even though, in general aluminum passivation requires a very small amount of oxidizers because the required aluminum oxide layer is very thin, whether or not the oxidizers in 1.4 ml IAS is enough to passivate a 6-ft aluminum thermosiphon is still unknown. The failure might be a result of a discontinuous liquid back flow, the lack of oxidizers, or both. In order to eliminate the effect of the lack of oxidizers, another 6-ft aluminum thermosiphon was charged with 2.6 ml of a three times concentrated IAS fluid. It means that there are 5.6 times more oxidizers as the previous charge. It has been demonstrated that the previous charge is more than enough to accomplish the passivation process of a 1-ft aluminum thermosiphon. Therefore, the new charge should have enough oxidizers to passivate a 6-ft aluminum thermosiphon. At least, the generation rate of NCG should be sharply reduced if lack of oxidizers is the dominant reason of the failure instead of the appearance of the discontinuous liquid back flow.

As seen in Fig. 16, 16 min after heating up, T1 started to decrease. It was a sign of the NCG appearance, and the time is about the same as the previous 6-ft thermosiphon test. With time lasting, more and more NCG was generated, and T2–T6 fell down one by one. It shows that, a lack of oxidizers is not the reason for the failure of the 6-ft aluminum thermosiphon, or at least it is not the dominant reason. Therefore, the discontinuous liquid back flow is a limit to be considered while using IAS fluid in aluminum-made phase-change heat transfer devices.

A continuous liquid back flow covering the whole internal surface is likely required while using IAS as the working fluid in aluminum phase-change heat transfer devices. It is a function of the tube geometries, the input power, and the charge amount. For the specific case investigated, the critical length was found to be between 1 and 2 ft. Adding grooves, screens, or sintered wicks to provide capillary paths could benefit the formation of a continuous liquid back flow.

Conclusion

It is demonstrated that IAS is compatible with aluminum phase-change heat transfer devices. When IAS is used as the working fluid in aluminum-made phase-change heat transfer devices, the oxidizers in IAS can passivate the aluminum surface by generating a thin and compact layer of aluminum oxide which protects the aluminum surface and prevents further reactions. In addition, an electrochemical circuit, formed by the body of the device and the liquid back flow, has been demonstrated to exist and thus enable the oxidizers in the evaporator to passivate the aluminum surface throughout the whole device without directly being in contact with it. Last, beside enough oxidizers are required to maintain the electrode potential and pH number of IAS in the passivation region of the Pourbaix diagram of aluminum during the lifetime, a continuous liquid back flow throughout the whole device has been demonstrated to be the key factor in maintaining the compatibility between IAS and aluminum-made phase-change heat transfer devices. An unstable or discontinuous liquid back flow will cause the generation and accumulation of hydrogen gas and eventually lead to the failure of the device. Adding a wick to help maintaining a continuous liquid flow during both storage and operation is important.

Future Work

First, it is mentioned that the passivation region of the Pourbaix diagram of aluminum moves with temperature changing. There will be a critical temperature that when the evaporating temperature is larger than it, the E and pH of IAS will locate outside of the passivation region. What the initial pH number of IAS should be in order to achieve the maximum critical temperature that requires further calculations and tests.

Second, the oxidizers in IAS are the key factor of the generation of aluminum oxide and the aluminum passivation. Permanganate is a stronger oxidizer than chromate, so it reacts with aluminum fast. It is used to form the initial aluminum oxide layer to protect the surface. However, permanganate is unstable, and an over amount of permanganate will lead to the generation of another NCG, oxygen gas. It is because that the over amount of permanganate will self-disassociate and release oxygen gas. Chromate, a slower reactor to aluminum, cannot passivate a large area of aluminum surface fast enough to prevent the generation of the hydrogen gas, so it only contributes to the healing of the aluminum oxide during the lifetime of operation. Therefore, further work is required to find the optimal amount of permanganate and the minimum amount of chromate for aluminum phase-change heat transfer devices. Both of them should be functions of the area of the total internal surface.

Third, chromium (VI) is carcinogenic, and it is accumulative in human's body. As a result, it is strictly restricted in some regions, such as Europe. Further researches are required to find another chemical constituent that can be used to replace the chromium (VI) in IAS. It should have a large solubility in water at a pH number range of 6–8. In addition, it should be an oxidizer having a similar oxidization ability with chromate at a pH number range of 6–8. When it reacts with metal surfaces, there should be no NCG generated as the product. Selenium (VI) might be a possible option.

Last, a continuous liquid flow is required to maintain the aluminum passivation during both storage and operation. For aluminum thermosiphons, further work is required to find the critical condensing heat flux, larger than which, a discontinuous liquid back flow can be avoided. In addition, adding a wick can help maintaining a continuous liquid flow. Further researches are required to find which type of wicks should be used and what material the wick should be made of.

Acknowledgment

We would like to acknowledge our colleagues: Sean Reilly, Jacob Supowit, and Ladan Amouzegar for their former work on IAS. In addition, we would like to acknowledge the support for this work under DARPA BAA 08-18 MACE program, DARPA BAA 07-36 TGP program, NSF IAS program, and AFRL Aluminum Passivation program.

This project was funded by Defense Advanced Research Projects Agency (DARPA) BAA 08-18 Micro-Technologies for Air-Cooled Exchangers (MACE) program, Award No. W31P4Q-09-1-005, Defense Advanced Research Projects Agency (DARPA) BAA 07-36 Thermal Grounding Plane (TGP) program, National Science Foundation (NSF) Inorganic Aqueous Solution (IAS) program, Award No. CBET-1336896, and Air Force Research Laboratory (AFRL) Aluminum Passivation program.

Funding Data

  • Air Force Research Laboratory.

  • Defense Advanced Research Projects Agency (Grant No. W31P4Q-09-1-005).

  • National Science Foundation (Grant No. CBET-1336896).

References

References
1.
Reay
,
D.
, and
Kew
,
P.
,
2006
,
Heat Pipes: Theory, Design, and Applications
,
5th ed.
,
Elsevier
,
Oxford
, UK.
2.
Kendig
,
M.
, and
Buchheit
,
R.
,
2003
, “
Corrosion Inhibition of Aluminum and Aluminum Alloys by Soluble Chromates, Chromate Coatings, and Chromate-Free Coatings
,”
Corrosion
,
59
(
5
), pp.
379
400
.
3.
Rocco
,
A. M.
,
Nogueira
,
T. M. C.
,
Simão
,
R. A.
, and
Lima
,
W. C.
,
2004
, “
Evaluation of Chromate Passivation and Chromate Conversion Coating on 55% Al–Zn Coated Steel
,”
Surf. Coat. Technol.
,
179
(
2–3
), pp.
135
144
.
4.
Reilly
,
S.
,
Amouzegar
,
L.
,
Tao
,
H. T.
, and
Catton
,
I.
,
2011
, “
Use of Inorganic Aqueous Solutions for Passivation of Heat Transfer Devices
,”
Tenth International Heat Pipe Symposium
(
IHPS
), Taipei, Taiwan, Nov. 6–9.
5.
Yao
,
Q.
,
2016
, “
Investigation of the Use of an Inorganic Aqueous Solution (IAS) in Phase Change Heat Transfer Devices
,”
Ph.D. dissertation
, University of California, Los Angeles, CA.
6.
Ahmad
,
Z.
,
2012
, Aluminium Alloys—New Trends in Fabrication and Applications, InTech, Rijeka, Croatia.
7.
Stubblebine
,
M. J.
,
Yao
,
Q.
,
Supowit
,
J.
, and
Catton
,
I.
,
2016
, “
A New Method for Evaluating Heat Pipe Fluid Compatibility
,”
Appl. Therm. Eng.
,
101
, pp.
796
803
.
8.
Ghiaasiaan
,
S. M.
,
2007
, Two-Phase Flow: Boiling and Condensation in Conventional and Miniature Systems, Cambridge University Press, Cambridge, UK.