The dual-temperature principle for providing reflux for the ammonia-hydrogen deuterium exchange process was proposed by the British firm Constructors John Brown [C12], has been tested in pilot-plant experiments conducted by Friedrich Uhde Gmbh at the plant of Farbwerke Hoechst in Germany [W2], and is to be used in a commercial plant at Talcher, India (item 19, Table 132), being constructed by Uhde.

Figure 13.37 is a material flow sheet for a dual-temperature ammonia-hydrogen exchange plant using the same amount of synthesis-gas feed and producing the same amount of enriched ammonia product as the monothermal ammonia-hydrogen exchange plant of Fig. 13.23. Comparison of these figures shows that the hot exchange column of Fig. 13.37 performs both the function of the ammonia dissociation step D of Fig. 13.23 in providing enriched

synthesis-gas recycle vapor for the enriching cold column of Fig. 13.37 and the function of the ammonia synthesis step A of Fig. 13.23 in providing depleted liquid ammonia reflux for the stripping cold column of Fig. 13.37. Comparison of these figures shows the advantages of the dual-temperature system to be as follows:

1. Elimination of the ammonia dissociation step D

2. Elimination of the work of recompressing synthesis gas from 55 to 350 atm

3. Elimination of the net heat input needed to dissociate ammonia at 740°C

4. Elimination of the need to synthesize ammonia for reflux and the costs associated with this step

5. Elimination of the catalyst deuterium stripper, F, Fig. 13.23

The dual-temperature system, however, is not without its disadvantages. Because the hot exchange column of Fig. 13.37 returns synthesis gas with a much lower deuterium content than the ammonia dissociation step of Fig. 13.23 and returns liquid ammonia reflux with a much higher deuterium content than the ammonia synthesis step of Fig. 13.23, it is necessary to operate the cold columns of Fig. 13.37 with higher liquid and vapor flow rates than those of Fig. 13.23 and to run them closer to minimum reflux conditions. Consequently, a much larger number of theoretical stages is needed in the cold columns of Fig. 13.37 than in the corresponding columns of Fig. 13.23. In addition, the dual-temperature system requires a large hot exchange column. Table 13.26 compares the liquid and vapor flow rates and number of theoretical stages in the two systems.

Even though flow conditions for the dual-temperature system, Fig. 13.37, were chosen to give a minimum number of stages, the increase from 5.7 stages for the monothermal system to

72.9 for the dual-temperature system must be viewed as a serious disadvantage of the latter. Another disadvantage of the dual-temperature flow sheet would be the complicated heat exchange system needed for heat recovery and humidification between the hot and cold towers, which is not shown in Fig. 13.37.

Nitschke [N2] has given a partial description of the flow sheet used by Uhde for the dual-temperature ammonia-hydrogen heavy-water plant that company is building at Talcher, India (item 19, Table 13.2). Figure 13.38 is a qualitative material flow sheet for the first-stage exchange columns of that plant.

Feed for this heavy-water plant consists of synthesis gas for the ammonia plant of the Indian Department of Atomic Energy, at 190 to 200 atm. The heavy-water plant, however, operates at 300 atm. To avoid the need for compressing synthesis gas, and to isolate gas flow in the ammonia plant from gas flow in the heavy-water plant, deuterium in feed synthesis gas is transferred to a solution of potassium amide in ammonia in the transfer column A, and synthesis gas 85 percent stripped of deuterium is returned to the ammonia plant.

Ammonia for the heavy-water plant is pumped to 300 atm, cooled to —25°C, and introduced as feed between the stripping (B) and enriching (C) sections of the first-stage cold exchange tower, where it joins ammonia circulating at the rate L In C the deuterium content of the ammonia is raised to first-stage product level xp by exchange against synthesis gas flowing at rate Gi + G2 whose deuterium content is reduced from yp to yp. A portion of the ammonia is sent to the first of two higher stages for further enrichment, and an equal flow of partially depleted ammonia is returned, reducing the deuterium content of ammonia entering the hot enriching section D to x’p. Here, because of the lower separation factor, the deuterium content of the ammonia is reduced to x’p, somewhat below that of feed, while the deuterium content of synthesis gas is raised from yF to yp. The deuterium content of ammonia is further reduced to the tails level xw in the hot stripping section E, where the gas flow rate has been reduced to Gi because of the recycle at rate Gi to the enriching sections C and D. The gas in E is enriched from yw to yp. A portion F of the tails is reenriched to feed level xF in the transfer column A, and the remainder, L, is fed to the cold stripping column В to be reenriched to feed level while stripping synthesis gas flowing at rate G from yF toy^.

The function of the four exchange-column sections can be better understood by reference

F+ L, Xyy

to the qualitative McCabe-Thiele diagram Fig. 13.39, whose nomenclature is keyed to Fig. 13.38. The slopes of the four operating lines are

Cold stripping, L/Gi

Cold enriching, CF + L)l(Gl + G2)

Hot enriching, (F + L)/(Gi + G2 )

Hot stripping, (F + L)/G1

By providing enough stripping plates, xw and у у/ could be made as close to zero as desired. By providing enough enriching plates, Xp and yp could be made as close to unity as desired.

7.2 Introduction

The possibility of using the slight differences that exist in the absorption spectra of isotopes of an element for isotope separation has been recognized ever since isotopes were discovered. The first reported successful photochemical separation of isotopes was that of Kuhn and Martin [K4], who dissociated C03sCl2 molecules in natural phosgene by light of 281.618-nm wavelength from an aluminum spark, which happened to be the correct wavelength. The first photochemical separation of isotopes on a practical scale was that of mercury isotopes. In one example [Zl], light from a mercury arc containing a preseparated mercury isotope was used to excite the same isotope in natural mercury vapor and cause it to form HgO with water vapor also present. Ibis method is not generally applicable to other elements because it makes use of the especially simple character of the mercury spectrum, with few, widely spaced lines.

Invention of the laser provided the intense, monochromatic, tunable light source needed to make photochemical isotope separation applicable to all elements, at least on a laboratory scale. The promise of this method was recognized as early as 1965 by Robieux and Auclair [Rl], who were issued the first patent on it. Since the pioneering experiments of Tiffany et al. [Tl] on bromine isotopes in 1966, an enormous amount of work has been done with lasers, with small-scale separation reported for most elements.

This text can describe only briefly the incomplete information publicly available on laser separation of uranium isotopes. For a more detailed discussion of the history and principles of laser isotope separation, reference may be made to the review articles of Letokhov and Moore [LI] and Aldridge et al. [A2], and to Farrar and Smith’s report on uranium [FI].

Two general methods have been proposed for separating uranium isotopes. In the photoionization method to be discussed in Sec. 9.2, MSU in uranium metal vapor is ionized selectively and then separated from unionized 23®U by deflection in electric or magnetic fields. In the photochemical method, to be described in Sec. 9.3, MSUFS in UF6 vapor is excited selectively and caused to react chemically to produce a solid lower fluoride, which is then separated from unreacted 23*UF6 vapor.

One type of tapered plant that is easy to treat theoretically, which has minimum interstage flow for a specified separation, and which is approximated by all isotope separation plants designed for minimum cost, is the so-called ideal cascade. An ideal cascade is one in which

1. The heads separation factor (1 is constant.

2. The heads stream and tails stream fed to each stage have the same composition:

хі+і=Уі-=2і (і = 2, 3, …,л —1)

The theory of such cascades was developed by P. A. M. Dirac and R. Peierls in England and by

K. Cohen and I. Kaplan in the United States and is described in The Theory of Isotope Separation by Cohen [C3]. The most important results are summarized in Secs. 8 through 12 of this chapter, with some changes in terminology and notation.

Deuterium was discovered by Urey et al. [U2] in samples of liquid hydrogen in which deuterium had been concentrated by partial evaporation. Because of the high deuterium separation factor, separation of deuterium by distillation of liquid hydrogen was studied by engineers in Germany [C3] and the United States [M8] during World War II and more recently by groups in the Soviet Union [Ml], France [Al], Germany [L2], Switzerland [H3], England [D2], and the United States [ВЗ, B4]. The main difficulties with the process have been the extremely low operating temperatures, which until recently have been without industrial precedent, and elimination of condensable impurities from the feed stream, which would foul heat exchangers and stop flow if not removed. Because these difficulties are those of low-temperature plants and are not unique to isotope separation, they will not be dealt with

extensively here. The references cited above may be consulted for more detailed information. Plants producing deuterium by distillation of liquid hydrogen that have been built and operated are listed in Table 13.7. The process used for the primary concentration of deuterium in all of these plants is similar in principle and is illustrated in generalized fashion in Fig. 13.5. The individual plants differ in detail; some of the principal differences are noted in Table 13.7. More detail is given in the references cited in Table 13.7.

The history of these plants has been sketched in Sec. 2.2 of Chap. 12. Each plant is parasitic to an ammonia synthesis plant, taking deuterium-bearing, hydrogen-rich feed gas from the ammonia plant, and returning gas depleted in deuterium to the ammonia plant, with little loss of hydrogen (less than 5 percent). The first two plants listed in Table 13.7 used as feed ammonia synthesis gas, which contains around 75 percent H2, 25 percent N2, and small amounts of CH4, A, C02, CO, 02, and H20. The remaining plants used as feed electrolytic hydrogen, which contains as impurities only H2 О and traces of N2 and 02. The high content of nitrogen and the presence of other impurities in the ammonia synthesis gas used as feed in the first two plants caused their design to be more complicated, their specific energy consumption higher, and the cost of heavy water produced in them greater than in the three plants using electrolytic hydrogen and feed. In fact, the first two plants were built primarily as pilot plants rather than as economic producers of heavy water, and they have been shut down, having served their purpose.

The process used in the primary section of these plants may be understood by reference to Fig. 13.1. Where gas from the ammonia plant is available under pressure, it is fed directly to the hydrogen distillation plant; otherwise it is compressed in the feed compressor. The gas is cooled down to around — 175°C by outflowing cold gas depleted in deuterium in a heat exchange system in which water is condensed and removed from the feed. Refrigeration to compensate for heat leaking into the plant may next be supplied to the feed. The gas is cooled further to about —245°C by outflowing cold gas in a second heat exchange system in which nitrogen is condensed and removed from the feed. Much nitrogen is condensed from synthesis gas; traces, from electrolytic hydrogen. Final cooling is provided by Joule-Thomson expansion through a valve, in which hydrogen is cooled to around —250°C and partially liquefied.

The hydrogen is distilled in the primary tower into a bottom product enriched in deuterium and an overhead product depleted in deuterium. Final concentration of the bottom product is effected by distillation either of liquid hydrogen or water (not shown in Fig. 13.1). The depleted hydrogen flows back through the feed exchanger system where it is warmed to room temperature. It is returned to the ammonia plant at the supply pressure, being compressed if necessary.

To provide heat to reboil the tower and to supply liquid hydrogen reflux, additional depleted hydrogen is circulated by the reflux compressor through another system of heat exchangers, to which additional refrigeration may be supplied. Cold, compressed hydrogen from this system flows through a coil at the bottom of the tower where it is condensed, supplying

Soviet Union Soviet govt.

9

?

Electrolytic Hj

4000

150

?

13

5000

2-4 6 + 70

9

9

Switch exchangers Adsorption

Liquid NH, + liquid N, Feed + recycle

Single

Bubble caps

1.05

77 plates 7-9

9

[Mil

Ems, Switzerland Sulzer Bros. Emswerke AG 1960 1967

Electrolytic H3

400

970

85

7

2400

3.7

14

1.5

7.5

Switch exchangers Switch exchangers Turbine expander Recycle

Single

Kuhn, Dixon 90 tubes, 5-cm diam. 2m 60

Water

[H31

Nangal, India Unde

Indian govt. 1962

Electrolytic H,

5300

450

90

45

2100

5

40

1

1.24

Regenerators Regenerators Uquid NH, + liquid N, Feed + recycle

Triple Sieve plates

9

9

4

Hydrogen

[Gil

heat to reboil the tower at the same time. This liquid is then expanded to tower pressure through a valve and fed to the top of the tower for reflux.

The product of these primary plants is a stream of hydrogen containing from 2 to 60 percent HD. At Ems this hydrogen was converted to water by burning it with oxygen, and pure heavy water was produced by distilling the water. At Toulouse, Hoechst, and Nangal, the HD-rich hydrogen stream was distilled directly to produce pure deuterium, which might then be burned to make heavy water. The basic flow sheet for this final concentration of deuterium was devised by Clusius and Starke [C7], who conducted the first experimental work on the fractional distillation of liquid hydrogen and showed that a mixture of hydrogen, HD, and deuterium could be separated by fractional distillation at atmospheric pressure into relatively pure fractions of H2, HD, and Ds without HD undergoing disproportionation and without appreciable conversion of ortho to para modifications.

The flow sheet for final concentration of deuterium developed by Clusius and Starke, which was used in the Hoechst and Nangal plants, is shown in Fig. 13.2, together with the primary tower of Fig. 13.1.

The bottoms from the primary tower are fed into the upper half of a smaller secondary tower, where fractionation into a bottom product of nearly pure HD is completed. This HD is warmed to room temperature in a heat exchanger and passed through a catalytic exchange reactor where its disproportionation into an equilibrium mixture of H2, HD, and D2 is catalyzed. The product of the exchange reaction is cooled to liquid hydrogen temperatures in the heat exchanger and fed to the bottom half of the secondary tower where it is fractionated into an overhead product of HD + H2 and a bottom product of pure deuterium. This is warmed to room temperature in the heat exchanger and constitutes the product of the plant. The HD and H2 overhead from the bottom of the secondary tower is fed to the top of the secondary tower for recovery of HD.

Heat to reboil these towers is provided by a stream of compressed, HD-free hydrogen,
which is condensed in reboiler coils located in the sump of these towers. The condensed HD-free hydrogen is then used as open reflux in the top of the primary tower. A Linde, double-column arrangement is used to provide reflux for the bottom of the secondary tower and reboil vapor for the top of this tower.

Of the plants listed in Table 13.7, the one at Nangal, India, may be regarded as indicating the full potentialities of this method of producing deuterium. It is a relatively large plant, producing around 141 of heavy water per year. It uses clean electrolytic hydrogen as feed. This hydrogen has been preconcentrated by two stages of partial electrolysis of water to around three times natural abundance. Power costs at Nangal, which is the site of a large hydroelectric project, are low. These three favorable circumstances make it possible to produce heavy water at a specific energy consumption in the distillation plant of only 2.1 kWh/g D20. This is lower than the energy consumption at the other sites, and of course is much lower than the 468 kWh/g D20 for electrolysis alone noted in Sec. 6. Garni et al. [Gl] in 1958 predicted that heavy water would be produced at Nangal at a cost of $27.2/lb or $60/kg. Data cited by these authors in 1958 as typical of what production rate and costs might be experienced at Nangal when the plant went into operation are summarized in Table 13.8.

A special problem of hydrogen distillation plants is the need to minimize conversion of ortho to para hydrogen. At room temperature, hydrogen contains 75 percent ortho and 25 percent para hydrogen. At low, hydrogen distillation temperature, the equilibrium proportion is nearly 100 percent para hydrogen. Conversion of ortho to para hydrogen is very slow in the absence of catalysts. Conversion must be minimized in a deuterium separation plant because about 1.5 times as much heat is released in conversion of ortho hydrogen as in liquefaction; it would greatly increase power consumption if allowed to occur. Conversion is catalyzed by paramagnetic materials, such as solid oxygen, and by ferromagnetic materials, such as certain steels. These must be excluded from the plant.

Condensed oxygen is especially objectionable, both because of the heat produced in ortho-para catalysis and because of its liability to explode when in contact with cold hydrogen.

Table 13.8 Production and cost data anticipated for Nangal heavy — water plant

Stages of electrolytic preconcentration, 2 Hydrogen production rate, 25,000 nm3/h Hydrogen feed rate to distillation plant, 5000 nm3/h Producing hours per year, 8000 D20 production rate, 14,000 kg/yr Erected cost of plant, $2.75 million

Partial descriptions of the type of equipment used in the gaseous diffusion plants of the U. S. DOE are given in references [Ul] and [Ш]. Figure 14.1 is a schematic plan view of three gaseous diffusion stages. The separating unit on each stage, called a converter, contains

thousands of tubes of diffusion barrier supported by tube sheets at each end. As UF6 gas at the highest process pressure flows along the inside of these tubes, about one-half of it effuses through the tubes into the region at the lowest process pressure outside of the tubes and is thereby slightly enriched in MSUF6. This low-pressure, slightly enriched gas, the stage heads stream, is compressed to an intermediate pressure in the first stage of a horizontally mounted, two-stage, axial-flow compressor of the next higher stage of the cascade. Here it is joined by an equal amount of UF6 at the same pressure and 23SU content representing the tails stream from the second higher stage of the cascade. The combined streams are compressed by the second stage of the compressor to the highest process pressure. The compressed gas flows through a cooler, where the heat of compression is removed by heat exchange against coolant C2F4C12, chosen because it will not react with UF6 should a leak occur. The compressed and cooled gas then flows through the tubes of the converter on the next higher stage of the cascade.

The tails stream from each converter, the gas that has not effused through the holes in the barrier tubes, flows through a control valve and into the intermediate pressure inlet of the compressor on the next lower stage of the cascade. The valve position is adjusted so as to control the pressure level of the converter upstream at the desired level.

In some stages of the U. S. plants the flow sequence is modified with the stage cooler inserted between the converter outlet for the heads stream and the compressor inlet. This permits the converter to operate at the compressor outlet temperature rather than the lower inlet temperature, and improves somewhat the separation performance of the barrier.

Figure 14.2 is a photograph of the process equipment used in the largest stages of the U. S.

Figure 14.3 Eurodif gaseous diffusion stage.

diffusion plants. The large drums in the foreground are the converters, each of which contains a cooler and thousands of barrier tubes. The two-stage axial-flow compressor that recompresses the UF6 that has passed through the barrier and circulates the undiffused gas is at the back of the figure. From 8 to 16 stages such as these are grouped into cells, housed in steel enclosures heated to around 60°C to prevent condensation of UF6. Each cell is the smallest independently operable unit, and is equipped with block and bypass valves to permit shutdown for maintenance.

As Fig. 12.2 showed, about 1270 stages are needed to separate natural uranium into product containing 3 w/o HSU and tails containing 0.2 w/o. The Portsmouth plant of U. S. DOE, which produces uranium enriched to 97 percent 235U, contains 4080 stages.

The large plants of U. S. DOE have operated for 20 years at a capacity factor over 99 percent and attest to the reliability of the gaseous diffusion process.

The fact that the total internal flow rate in a close-separation, ideal cascade is given by Eq. (12.142) may be derived without solving explicitly for the individual internal flow rates by the following development, due originally to P. A. M. Dirac. This procedure is valuable in showing the fundamental character of the separation potential and the separative capacity, and provides a point of departure for the treatment of multicomponent isotope separation.

We consider a close-separation, ideal cascade whose external streams have molar flow rates Xk (positive if a product, negative if a feed), and compositions xk expressed as mole fraction. Let us look for a function of composition ф(хк), to be called the separation potential, with the property that the sum over all external streams, to be called the separative capacity D,

D = 2 Хкф(хк) (12.162)

is proportional to the sum of the flow rate of all internal streams. At this point in the derivation, the nature of ф(хк) is assumed not to be known.

Figure 12.22 represents stages i— 2, і — 1, i, and / + 1 of such a cascade, with the fcth product stream consisting of part of the heads stream of stage і — 1. The total internal flow from stage і is Mt + JV). The separative capacity of the r’th stage, considered as an isolated plant, is

Ді = міФ<Уі) + Nttfxt) — {Mt + NtWzt) = {Mt + ЮІЄіФі) + (1 — віУр(Хі) — фі)]

and

Substitution of these expansions into (12.163) yields

Д/ = [8 i(yі — г,)2 + (1 — віфі — z, f]

where the term in d<j>ldz has dropped out because of the material-balance relations (12.8) and (12.9).

In a close-separation cascade,

Уі ~ z, = (1 — 0/X« — l)z,(l — Zi) (12.167)

as may be seen from (12.18) and (12.21), with (a — 1) and (j3 — 1) considered small relative to unity. Similarly,

Xi — Zi = — в,{a — l)z,-(l — z,)

Figure 12.22 Flow in portion of ideal cascade. Molar flow rates denoted by capital letters, mole fractions by small letters.

(12.173)

where Д,- is defined by (12.163).

When the separation potential satisfies (12.172), the separative capacity of a single stage in a close-separation cascade operated at a cut of ^(Af = iV) from Eq. (12.170) is

_ M,(a — If

‘І——- A——

We shall now show that the separative capacity of the entire cascade, D, is given by

an

all external

stages streams

D= I A’= £ ХкФ&к) (12.175)

Consider first the sum of the separative capacity of stages і and / — 1.

Ді + ін = + Nttix,) — + NfWz,) +

+ +^нЖгн) 02.176)

The internal streams between this pair of stages, Af,_ t and Nt, may be expressed in terms of

the streams external to this pair of stages Mit Ni+U Xk, Af,_2, and 7V,_ j by means of the

material-balance relations:

Nt = Ni-y. — M,_2 (12.177)

and Nt-Mi-l=Ni+i-M,-Xk (12.178)

Because of the assumption that this is an ideal cascade,

By means of these four equations (12.176) may be expressed as

Д/ + At — і = Міф(уі) — Мі+1ф(хі+1) — М^Фі-т) + + Xкф(хк) (12.181)

This is an example of (12.175) applied to the pair of stages і and і— 1. If Д<+1 is added to this expression, terms in M1 and Nl+, may be eliminated in the same way. By proceeding in this way until the separative capacity of every stage has been included in the sum, terms representing all internal streams cancel out, the only terms that remain on the right are those representing external streams, and Eq. (12.175) results.

Thus, we have shown that the separative capacity of an ideal cascade is the sum of the separative capacities of its component stages. And if the separation potential satisfies the differential equation (12.172), the total internal flow is given by

external

streams

j+K=j^y Y, Xk<Kxk) (12-182)

as was to have been shown.

The general solution of (12.172) is

Ф = (2x — 1) In — p*— + ax + b (12.183)

Here a and b are arbitrary constants, and the general composition variable x has been substituted for Zj. The arbitrary constants a and b do not affect the value of the right side of (12.182) because of the overall material-balance relations

£*fcxfc=0 (12.184)

к

and £ЛГ*=° (12.185)

к

In Eq. (12.157) for the price of uranium, it may be noted that the term in brackets has the general form (12.183) for the separation potential, with

a = 219.5666 (12.186)

and b = —6.4300 (12.187)

The separation potential may be thought of as related to the value of a mixture of isotopes, and has, in fact, been called the “value function” by Cohen [СЗ].

In addition to needing an exchange equilibrium constant different from unity, exchange processes for concentrating deuterium require a reflux-making step in which part of the deuterium in the liquid phase leaving an enriching column is transferred to the vapor phase returned to the column. This can be done either by a chemical-phase conversion operation in the monothermal exchange processes to be described in this section or by another exchange column at a higher temperature in the dual-temperature exchange processes to be described in Secs. 11 through 14.

The exchange towers of Fig. 13.21 are an example of a monothermal exchange process for concentrating deuterium, with the electrolytic cells providing reflux-making phase conversion. Because the equilibrium constant for the reaction

HD(?) + H2 0(0 — H2 (g) + HDO(/)

is 3.2 at 60°C, a flow sheet similar to this figure would permit concentration of deuterium even if no separation occurred in electrolysis.

Chemical reflux-making steps, such as the electrolysis of water in Fig. 13.21, cost more than thermal reflux-making steps such as the reboiler or condenser in distillation. Consequently, there are only a few examples of economical monothermal exchange processes for concentrating deuterium. Hydrogen-water exchange refluxed by water electrolysis as in Fig. 13.21 is one example that can be economical where electricity is cheap enough and electrolytic hydrogen valuable. The only other commercial example of production of deuterium by monothermal exchange is use of the ammonia-hydrogen exchange reaction

HDfe) + NH3(0 — H2 fe) + NH2 Щ)

This was used m the exchange plant at Mazingarbe (item 12, Table 13.2) and is planned in plants in India (items 16 and 18). The economic attractiveness of this process comes from two factors:

1. The relative ease with which hydrogen reflux can be obtained by thermal cracking of ammonia

2. The relatively low cost of providing liquid ammonia reflux as incremental production of an existing synthetic ammonia plant

10.1 Monothermal Ammonia-Hydrogen Exchange

Figure 13.23 illustrates the principle of enriching deuterium by monothermal ammonia — hydrogen exchange in conjunction with a synthetic ammonia plant. The flow quantities have been developed from partial information reported for the Mazingarbe plant [LI, El, N2]. Feed, at point (1), consists of ammonia synthesis gas, 3H2/1N2, which has been purified to reduce its content of water, CO, and C02 to less than 1 ppm and compressed to the pressure of ammonia

Figure 13.23 Monothermal ammonia-hydrogen exchange process. Flow quantities, kg-mol/h.

synthesis, 350 atiri^ in this example, as at Mazingarbe. At (2) feed joins additional synthesis gas circulating countercurrent to liquid ammonia in the stripping exchange column В and the enriching exchange column C, both operated at —25°C. At this temperature, the equilibrium constant for the foregoing deuterium exchange reaction is 5.2. At the countercurrent flow conditions of Fig. 13.23, deuterium is transferred from upflowing synthesis gas to downflowing ammonia. For the reported [El] 85 percent recovery of deuterium, the atom percent deuterium in synthesis gas is reduced from 0.0132 percent in feed [LI] to 0.00198 percent in the gas leaving the stripping column (4). For the reported [LI] 100-fold enrichment, the ammonia leaving the enriching column (11) is enriched to 1.32 percent deuterium. Final concentration to 99.8 percent deuterium [N2] is by distillation of ammonia, G.

At Mazingarbe synthesis gas leaving the stripping column at (4) was converted to ammonia in the synthetic ammonia plant A of Houffleres du Bassin du Nord et du Pas-de-Calais. In Fig. 13.23 about 5 percent of the synthesis gas is purged at (7) to remove inert impurities, mostly methane and argon, present in feed. About 65 percent of the ammonia synthesized in A is withdrawn as stripped ammonia product at (6) and the remainder is returned (8) to the exchange column as liquid reflux.

At the bottom of the enriching column C, a small portion of the partially enriched ammonia (13) is withdrawn for final concentration, and the remainder (19) is dissociated in D into partially enriched synthesis gas (12) to serve as reboil vapor for the enriching column. To obtain nearly complete dissociation of ammonia without using excessively high temperatures, the Mazingarbe plant is reported [LI] to have reduced the ammonia pressure in the cracking section to 55 atm. At this pressure and the assumed cracking temperature of 740°C, the ammonia content at equilibrium would be under 1 percent. For simplicity, the small amounts of ammonia in recycle gas (12) and hydrogen dissolved in recycle ammonia (3) have been neglected in Fig. 13.23.

To cause the deuterium exchange reaction between hydrogen and ammonia to take place at a useful rate, it is necessary to have 1 to 2 m/o (mole percent) of potassium amide, KNH2, dissolved in the liquid phase to serve as a homogenous catalyst [C2]. Presence of potassium amide complicates the process for three reasons:

1. Potassium amide is expensive and must be recovered and recycled.

2. It remains in deuterium exchange equilibrium with ammonia:

KNHD + NH3 ^ KNH2 + NH2D

3. It reacts, sometimes violently, with oxygen-containing impurities in synthesis gas.

To recover and recycle potassium amide, the partially enriched ammonia containing dissolved partially enriched potassium amide leaving the exchange column at (11) is depressured to 55 atm, heated to 125°C, and separated at E into ammonia vapor and a concentrated liquid solution of potassium amide (16). Because this solution contains 1.32 percent deuterium, its deuterium content must be reduced before it is recycled to the top of the stripping exchange column. In Fig. 13.23 this is done by countercurrent exchange with stripped ammonia vapor (9) in the catalyst deuterium stripper F. This is a conventional sieve-plate column in which deuterium is transferred from dissolved potassium amide to ammonia vapor in the overall reaction

KNHD(0 + NH3(?) — KNH2(/) + NHj D(g)

which proceeds rapidly with an equilibrium constant near unity. The deuterium-depleted catalyst solution leaving F is repressured to 350 atm and returned (10) to the top of the

^l atm = 1.01325 bar.

stripping exchange column B. At Mazingarbe, catalyst stripping was done with synthesis gas feed.

Potassium amide reacts with water, C02, CO, and oxygen, forming solid impurities that would plug the columns. With oxygen it forms potentially explosive potassium azide, KN3. To prevent loss of exchange catalyst and formation of undesirable reaction products in the exchange system, synthesis gas feed is purified ahead of (1) by two systems not shown. It is dried by molecular sieves and then passed through a guard chamber containing sacrificial KNH2 dissolved in liquid ammonia to remove oxygen-containing impurities [N2].

The flow quantities of synthesis gas feed (I), stripped ammonia product (6), and enriched ammonia product (14) were calculated from the net ammonia production rate of 330 MT/day [El] reported at Mazingarbe, the reported [N2] heavy-water production rate of 26 MT of 99.8 percent D20 per year at 85 percent recovery [El], and the 0.0132 percent deuterium in feed [LI]. The reflux rates of ammonia (3) to column В and synthesis gas (12) to column C and the number of theoretical stages were evaluated in unpublished design studies at Massachusetts Institute of Technology which led to a requirement of n$ = 2.6 theoretical stages in stripping column В and ri£ = 3.1 theoretical stages in enriching column C. These values were obtained from the Kremser equation (13.92). The number of stages in the stripping section, п$, is

In [(ay2 — x3)/(oy4 — x5)]

In (oL3/Kj)

The number of stages in the enriching section, ng, is

In [(oy12 -*n)/(oy2 ~x3)] "£ hi (oLu/KI2)

In these equations у is the atom fraction deuterium in hydrogen, and x is the atom fraction deuterium in the solution of potassium amide in ammonia at the numbered points in Fig. 13.23. К is the molar flow rate of hydrogen in the vapor phase, and L is the molar flow rate of hydrogen in the liquid solution of KNH2 in ammonia. For example, L3 = f(438.80) + 6.20 = 664.4.

The small number of theoretical stages is a consequence of the high value of the separation factor, 5.2, and is the principal advantage of the monothermal ammonia-hydrogen exchange process. There are, however, a number of offsetting disadvantages. Even in the presence of KNH2 catalyst, the rate of the exchange reaction is low, primarily because of the low solubility of hydrogen in liquid ammonia. Even at the high pressure of 350 atm used in Fig. 13.23, the hydrogen content of the liquid is only 0.85 m/o. With conventional sieve-plate or bubble-plate columns the plate efficiency would be only 1 or 2 percent, necessitating use of hundreds of plates. The Mazingarbe plant is reported [LI] to have used special ejectors for upflow of gas to entrain liquid and increase interphase transfer area. Because of the pressure drop taken by the upflowing gas, it was necessary to pump the liquid from one plate to the plate next below. Even with this enhanced contacting, it was necessary to use towers 36 m high [El].

The Mazingarbe exchange plant produced its first heavy water in January 1968. It was taken out of service in 1972 because of an explosion in the ammonia synthesis plant, which has not been repaired. Operation of the exchange plant itself was satisfactory; the availability factor was 92 percent in 1970. Lefrancois stated [LI] that an operating temperature of —10°C in the exchange towers would have been preferable to the design temperature of —25°C.

Evolution of process. The separation nozzle process has evolved through a number of forms. The first process tested experimentally by Becker [B6] is illustrated schematically in Fig. 14.20, with dimensions for one of the devices tested on UF6. UF6 feed at a pressure p of around 20 Torr flows through a slit-shaped nozzle 0.045 mm wide into a region at much lower pressure p’, where a fraction в, about 0.2, of the feed diverges from the feed jet and is somewhat enriched in the light isotope. The remaining fraction of the feed jet, 1—6, somewhat enriched in the heavy isotope, passes through a wider separator slit, where its pressure p" is somewhat higher than p’ because of deceleration. London [L4] gives examples of the separation factor, cut, and UF6 feed rate observed by Becker [B6]. Optimum pressure conditions at which power consumption, compressor capacity, and nozzle length per unit separative capacity were smallest are listed in the first column of Table 14.18, together with the minimum values of these performance indices.

Comparison with corresponding performance indices for gaseous diffusion, taken from

Table 14.9, shows that in this early version of the separation nozzle process, the separation factor was slightly better than for gaseous diffusion, but the power consumption, Q/Д, the rate of loss of availability, was four times as great as in gaseous diffusion, and the compressor capacity, V/A, was ISO times as great. The high power consumption was a consequence of the high pressure ratio through which both light and heavy fractions were expanded in this early version of the nozzle process, and the very high compressor capacity was caused both by the high pressure ratio and the low operating pressure level.

Two modifications of the process developed by Becker and his associates have greatly improved these process characteristics. (1) Dilution of UF6 feed with a gas of lower molecular weight, helium in early developments [B7] and hydrogen in later developments [Gl], has had two beneficial effects. Sonic velocity in the nozzle is increased, with accompanying increase in separation factor, and diffusion rates are increased, permitting operation at higher pressure and higher uranium throughout without impairment of separation. (2) The radical change in nozzle geometry illustrated in Fig. 14.21 adds the relatively large separation caused by centrifugal acceleration to the smaller separation accompanying expansion through the slit.

Improved nozzle process. In Fig. 14.21, a dilute mixture of / mole fraction UF6 in hydrogen at upstream pressure p is expanded through a convergent-divergent slit with a throat spacing s into a curved groove of radius a. After being deflected through 180° by the wall of the curved groove, the gas stream at lower pressure p traveling at high speed is separated by a flow divider set at radius c into an outer heavy fraction depleted in 23SUF6 and hydrogen and an inner light fraction enriched in these components. The cut в is determined by the position of the flow divider. The separation factor a (1) is higher the higher the speed attained by the gas, which is higher the higher the pressure ratio р/р’ and the lower the UF6 content of the feed gas; (2) has a maximum value at an optimum pressure level, which is inversely proportional to the dimensions s and a; and (3) is higher the lower the cut в.

Figure 14.22 shows the dependence of separation factor on cut. The lower lines show the separation factor calculated by assuming that the 23SUF6 and 238UF6 density distribution in the curved groove reaches centrifugal equilibrium at the indicated peripheral speed v, using the theory to be derived later in this section. The top line shows the highest values of the separation factor reported in Becker’s papers [BIO], at a pressure ratio of 8 and a low UF6 con­tent, 1.6 m/o (mole percent) in hydrogen, at which the calculated outlet gas velocity for reversible expansion is 1042 m/s. Because these extreme conditions result in gas-compression energy consumption per unit of separative work produced much greater than optimum, they are not recommended for a commercial plant. They do indicate, however, that values of a — 1 in the current version of the nozzle process can be 10 times as high as in the early, linear nozzle process of Fig. 14.20 or in the gaseous diffusion process of Table 14.9. ■

Design studies for a commercial plant by Geppert and associates [Gl] indicate that optimum conditions are feed composition f = 0.042 mole fraction UF6 in hydrogen, pressure ratio р/р = 2.1, and a cut в = at which a — 1 = 0.0148, still four times that in gaseous diffusion, and somewhat higher than what would be predicted for centrifugal equilibrium at the speed attainable from expansion through this pressure ratio. The cut of j necessitates use of a three-up, one-down cascade, as shown in Sec. 14.2 of Chap. 12.

Attainment of separation factors higher than predicted for equilibrium in a centrifugal field have been explained by Becker and associates [B10] as follows. Before the mixture of hydrogen, 235UF6, and 238UF6 enters the curved groove, the concentration of each is spatially uniform. While undergoing linear and centrifugal acceleration, the heaviest component, 238UF6, experiences the highest forces and migrates more rapidly toward the outer wall than the lighter component, 235UF6. Thus, there is a transient time during flow along the curved wall when the 238UF6/23sUF6 concentration ratio is a maximum, after which the ratio decreases toward the limiting, equilibrium value. This transient phenomenon is enhanced by high dilution by hydrogen, which reduces the frequency of collisions between 235UF6 and 238UF6 molecules, which otherwise would speed attainment of centrifugal equilibrium between these species. Maximum benefit from this transient phenomenon for a given pressure ratio is obtained at an optimum pressure level for a given set of nozzle dimensions. At a pressure level lower than

optimum, diffusion rates, which are inversely proportional to pressure, cause attainment of centrifugal equilibrium before the gas mixture reaches the flow divider. At a pressure level higher than optimum, diffusion rates are too slow to permit the initially spatially uniform 238UF6/235UF6 ratio to reach its maximum transient value.

The left half of Fig. 14.23 shows the dependence of separation factor, expressed as a — 1, on pressure ratio р/р’ and upstream pressure p, for a cut в = 3 and / = 0.04 mole fraction UF6 in hydrogen, as reported by Becker et al. [В10]. At each pressure ratio р/р’ there is an

Power per unit

Seporotion factor separative capacity, Q/й

Upstream pressure, p, Ton-

optimum inlet pressure p at which the separation factor is a maximum. At high pressure ratios, the separation factor is higher than predicted by Fig. 14.22 for centrifugal equilibrium at a cut of 3, for any speed. At each pressure ratio, i. e., at each speed, there is an inlet pressure at which the separation factor is a maximum; this inlet pressure is higher the higher the pressure ratio and the higher the speed.

The right half of Fig. 14.23 shows the dependence of power consumption per unit separative capacity Q/A on the same pressure variables. The power consumption has been calculated as the rate of loss of availability, so that Q/A is given by

Q 2RT0]np/p’

Д ~ (a — l)20(1 — в)

Optimum pressure conditions for minimum Q/A are inlet pressure p = 22 Torr, and pressure ratio р/р = 2.1, at which a — 1 = 0.0148 and the power consumption is 0.31 kW/(kg SWU/year).

The nozzle dimensions a and s with which the pressure level of Fig. 14.23 was associated were not stated in reference [В10]. Dimensions and related operating pressures reported [VI] as optimum for UF6-helium mixtures are

Because the diffusion coefficient of UF6 into hydrogen is about 20 percent higher than into helium, optimum pressures for UF6-hydrogen mixtures would be about 20 percent higher than the foregoing values. The inference then is that the data of Fig. 14.23 were obtained with a nozzle with a throat spacing around 0.4 mm.

Operation at the highest feasible pressure is economically desirable because the volumetric flow rate is lower and compressors and piping are smaller. Later design studies for a commercial plant by Geppert et al. [Gl] selected optimum outlet and inlet pressures of 138 and 290 Torr, respectively. These are for 4.2 m/o UF6 in hydrogen feed, presumably with nozzle dimensions of

Throat spacing s = 0.03 mm Groove radius a = 0.1 mm

the smallest dimensions reported [VI]. This lower pressure ratio of 2.1 was chosen to reduce the specific power consumption and to permit operation with a single stage of compression without intercooling.

The second column of Table 14.18 summarizes characteristics of the improved nozzle plant whose design was described by Geppert et al. [Gl], The slit length, power, and compressor capacity per unit separative capacity are greatly improved over the early process because of the much higher separation factor and operating pressures. However, the last two are still not as small as those for the gaseous diffusion process, restated from Table 14.9 in the third column. The higher compressor capacity and power consumption of the nozzle process compared with gaseous diffusion results from the 24-fold dilution of UF6 with hydrogen and the need to recompress both light and heavy fractions through the full pressure ratio in the nozzle process. However, the much higher separation factor of the nozzle process causes the number of stages it requires to be only 40 percent of those needed by gaseous diffusion for the same separation,
despite the smaller cut used in the nozzle process. When all sources of process inefficiency, such as pressure drops and compressor inefficiency, are taken into account, Geppert [Gl] has estimated that the actual power consumption of a complete nozzle plant with capacity of

5,045,0 kg SWU/уеаг would be 2520 MW, for a specific power consumption of 0.50 kW/(kg SWU/уеаг). This may be compared with the capacity of the gaseous diffusion plants of the U. S. DOE, 17,230,000 kg SWU/уеаг and their power consumption of 6,060 MW, for a specific power consumption of 0.352 kW/(kg SWU/уеаг). These actual power consumptions are in approximately the same ratio as the values of Q/A in Table 14.18.

Equipment of nozzle plants. Becker [Bll] has described two types of separating elements with the cross-section contour shown in Fig. 14.21. The more fully developed type, produced by mechanical means by Messerschmidt-Bolkow-Blohm Gmbh, Munich, is illustrated in Fig. 14.24. This consists of a cylindrical aluminum tube 2 m long, whose outer surface carries 10 semicircular longitudinal grooves, through each of which a portion of the feed gas flows circumferentially. The convergent-divergent nozzle contour and flow divider are provided by properly shaped strips fitted into 10 dovetail-shaped notches cut into the aluminium tube. The aluminum tube is divided into 10 radial sectors which carry, alternately, inflowing feed gas and outflowing heavy fraction. The light fraction flows into the space outside the tube through a slot between the dovetail strips, which are held in position by small spherical spacers at regular intervals. A complete separation stage contains 80 or more of these separating tubes mounted vertically, with appropriate headers for admitting feed and withdrawing light and heavy fractions. Their predicted separating capacity when operated on 4.2 percent UF6 in hydrogen and pressures of 290 and 138 Torr is 0.48 kg SWU per year per meter slit length [Gl]. A separating element of this type has been run on UF6 for over 30,000 h without change in measured separation factor [В11]. The cost of mass-produced separating tubes of this type predicted in 1971 [B9] was less than $16/(kg SWU/year).

A second type of separating element, developed by Siemens AG, is fabricated by photoetching of metal foils by techniques used in miniaturizing electronic circuits. The left side of Fig. 14.25 is an enlarged contact print of such an etched foil. The middle of Fig. 14.25 shows how these foils are stacked into chips held by cover plates pierced with holes in register with the feed and heavy fraction passages. The right side of Fig. 14.25 shows assembly of chips into a tube.

Figure 14.26 is a partially cutaway side view of a small prototype separation nozzle stage that has been run [Gl] on total recycle with UF6 and hydrogen. The stage contains 54 of the 10-sector elements 1 m long. The separating elements are mounted vertically inside a metal tank from which is suspended a two-stage gas cooler and a two-stage radial centrifugal compressor. The two-stage arrangement was necessitated by design for a compression ratio of 4. Stages for a larger production plant, based on later designs, will use a compression ratio of 2.1, and a single-stage cooler and axial-flow compressor.

Theory of separation. Theoretical analysis of the current form of the separation nozzle process is very difficult because of the presence of three components of widely different molecular weight, the complex flow geometry, and the importance of transient diffusion effects during the brief exposure of the mixture to centrifugal acceleration. A simplified, approximate analysis of the effect of cut and gas velocity on separation factor, separative capacity, and power consumption will be given by assuming (1) that 23SUF6 and 238UF6 attain their equilibrium concentration distribution at the end of the 180° rotation the expanded gas undergoes, and (2) that gas motion is in “wheel flow” at uniform angular velocity со. Finally, the effect of factors neglected in this simplified treatment will be discussed qualitatively. Mailing and Von Halle [М2] made similar assumptions in their simplified analysis of the nozzle process.

The flow geometry assumed is illustrated in Fig. 14.21. The gas mixture is assumed to be rotating at uniform angular velocity со in a semicircular groove of radius a. Centrifugal equilibrium is established where the mixture is separated by the flow divider at radius c into an inner, light fraction enriched in hydrogen and 23SUF6 and an outer, heavy fraction depleted in these components relative to 238UF6.

From the treatment of the gas centrifuge in Sec. 5.5, the dependence of concentration of

light isotope (e. g., 23SUF6) on radius r at centrifugal equilibrium is

where Pi(0) is the density of component (1) of molecular weight m, at the center of rotation (r = 0). A similar equation for the density of component 2 (e. g., J38UF6) is

(14.265)

Ґ is the absolute temperature of the mixture after acceleration to angular velocity u>. If the flow divider is set at r = c, the mass flow rate of component 1 in the light fraction per unit length is

Ґ fc /m1wJr! RT’pM

= u>rpi(r) dr = / ojrpі (0) exp I dr =————-

Jo Jo 2RT J m, w

(14.266)

Similarly, the mass flow rate per unit length of component 2 in the light fraction is fc RT’pi(0) Г /т2ш2<^ "І

=l wrp’(r)dr = _^T (14-267)

The mass flow rate of component 1 in the heavy fraction flowing between radius c and the outer wall at radius a is

and that of component 2 in the heavy fraction is

Let

(14.270)

as in Sec. 5.5. In the low-enrichment case, when зЯі < 3R2 and 3t, < , the cut в is

_ _JKj______ exp (A2c2/a2) — 1

зи2 + aij exp A2 — 1

Hence, the fraction of the total flow area used for the light fraction to provide a cut of в is

2 = 1 + ~ In [в + (1 — в) exp (-Л5)]

The fraction of the flow area used for the light fraction has a lower limit of в when the speed is low (A -*■ 0) and approaches unity as the speed increases (A -* °°), as in the countercurrent centrifuge, because all flow is compressed against the outer wall.

The separation factor a is

This notation is used to facilitate comparison with gaseous diffusion, for which the ideal separa­tion factor is

(14.275)

With cJ/e2 from (14.272),

1—6 ([6 + (1 — 0) exp (—A2)1>a’ — exp (—A3/aJ) в / l-[6 + (l-0) exp(-A2)],/a5

At low speed (A -*■ 0), a approaches unity. At high speed,

1-е Є*/“«

When a0 — 1 < 1, as in uranium isotope separation,

The corresponding expression for a cross-flow gaseous diffusion stage, from Eqs. (14.92) and (14.93), is

, («о “ ln(l — Є)

<“ — Ода =———————- fl—————

Hence, in the nozzle process at high speed, the separation factor at cut Є is 2ШмЕв times as great as in gaseous diffusion at cut 1 — Є.

In Fig. 14.22 the curves of separation factor versus cut for centrifugal equilibrium were calculated from Eq. (14.276) for 235UF6 (m, = 349) and 238UF6 (m2 = 352). oj = 1.008596.

The temperature Ґ and peripheral speed coa = v occurring in the definition of A2, Eq. (14.270), are for the mixture of UF6 and hydrogen after expansion to speed v. The nozzle process ordinarily is operated at a known constant temperature T before expansion. T, Ґ, and v are related by the enthalpy balance

CP(T (14.280)

where Cp is the molar specific heat at constant pressure and m is the molecular weight. At T = 313 K, assumed [Gl] as the temperature at which the mixture of UF6 and H2 enters the nozzle separator,

Cp{H2) = 6.874 cal/(g-mol-K) [P2]

CP(UF6) = 31.3 cal/(g-mol-K) [D6]

and Cp(mixture) = 6.874(1 — f) + 31.3/ cal/(g-mol • K) (14.281)

where / is the mole fraction of UF6. In dealing with gas expansion processes, it is conventional to use the heat capacity ratio

—4-

Cv 1 — R/Cp

The molecular weight m of a mixture of UF6 and hydrogen is m = 2.016(1 — Л + 352.02/

On the assumption that и = сое, from Eqs. (14.270), (14.280), and (14.282),

Values of и calculated from Eq. (14.284) for the values of A2 shown in Fig. 14.22 are tabulated at the bottom of Fig. 14.27 for several mole fractions of UF6,/. The equilibrium separation factor increases rapidly between 100 and 250 m/s and approaches a limiting value

A 9(1-9X<*-1)2 Z 2

from Eqs. (12.169) and (12.172). Figure 14.27 shows the dependence of this separative capacity on cut for the peripheral speeds used in Fig. 14.22. The important point to note is that the cut at which separative capacity is highest for a given speed shifts from в = 5 at low speed to в = I at the highest speeds. Because both the light and heavy fractions have to be recompressed in this version of the nozzle process, the cut at which the separative capacity is highest is the cut at which power consumption is lowest for a given speed.

Power requirement. In the separation nozzle elements shown in Figs. 14.24 and 14.25, the kinetic energy of the expanded gas is dissipated after separation. Then the minimum net power to recompress the gases leaving the separator at pressure p to the feed pressure p is

ZRTp In (р/р1) 238/

The minimum power consumption per unit separative capacity is obtained from (14.290) and (14.285):

The dependence of a on A2 and 0 is given by (14.276).

For every feed composition / and cut 0, there will be an optimum value of ^42, because the numerator of (14.291) increases continuously with A2, whereas the denominator approaches a limit. As a practical matter, values of A2 are limited to those corresponding to the speed of sound because expansion through the curved nozzle becomes very irreversible at higher speeds. Because the sonic speed is

(14.292)

(14.293)

(14.294)

The lower curve of Fig. 14.28 is a plot of (Q/A)s versus mole fraction UF6 in feed,/, for 0 =

5. The minimum value of (Q/A)s is 0.072 kW/(kg SWU/уеаг) at a feed composition of 0.18 mole fraction UF6. This is to be contrasted with the optimum value of 0.31 kW/(kg SWU/year) reported by Geppert et al. [Gl] for experiments with a feed composition of 0.04 mole fraction UF6, and a design value of 0.287 kW/(kg SWU/year) for a commercial plant with a feed composition of 0.042 mole fraction (Table 14.18).

Part of the lack of agreement can be explained by the fact that flow in the curved groove in which separation takes place is quite different from the wheel flow assumed in the foregoing derivation. Instead of v at the wall (r = a) being a maximum as assumed, v actually drops to zero there because of wall friction. Also, flow through the curved nozzle cannot be perfectly reversible, so that the speed of the mixture after expansion will be lower than calculated for reversible expansion through a given pressure ratio. Justification for the choice of a feed

composition of 0.042 fraction UF6 and agreement with Geppert’s reported Q/A of 0.31 kW/(kg SWU/уеаг) can be obtained by assuming that the effective peripheral speed и of the gas after expansion through the pressure ratio corresponding to sonic speed is one-half the sonic speed. The upper curve of Fig. 14.28 was calculated for this condition. The minimum value of 0.308 at a feed composition of 0.042 mole fraction UF6 in hydrogen is close to the values cited by Geppert [Gl].

One of the most compact and efficient of the mixer-settlers is the pump-mix mixer-settler, developed by Coplan et al. [C5, C6] specifically for radiochemical separations. One stage of this device is shown schematically in Fig. 4.22. A countercurrent cascade of mixer-settlers is shown in Fig. 4.23. The stage consists of a mixing chamber at the left of Fig. 4.22 and a settling chamber at the right. The rotating impeller in the mixing chamber serves to promote equilibrium contact between light and heavy phases, to pump phases between adjacent stages, and to control liquid levels. The mixing chamber is divided into an upper and a lower compartment by a horizontal baffle, pierced with a hole somewhat larger than the impeller shaft. This shaft, which is hollow, passes through the upper compartment of the mixing chamber and dips into the lower. The impeller draws liquid from the bottom compartment through the hollow shaft and discharges it through holes between the blades into the upper compartment. This pumping action of the impeller maintains an interface between mixed phases and the heavy phase at the bottom of the shaft. A free surface between mixed phases and air is maintained in the upper part of the mixing chamber. This free surface is progressively lower in adjacent stages in the direction of light-phase flow.

Mixed phases flow by gravity from the mixing chamber past a baffle into the settling chamber (closed at top) where the two phases separate. The heavy phase flows through a trapped outlet near the bottom (not shown) into the bottom compartment of the mixing chamber of the adjacent stage, where the interface controlled by the stirrer shaft is at a lower level. The light phase flows through an outlet near the top of the settling chamber directly into the mixing chamber of the adjacent stage in the opposite direction.

Flow of heavy phase in one direction is induced by the pumping action of the stirrer, coupled with the trap between settling and mixing sections; flow of light phase in the opposite direction is induced by the progressively lower level of the free surface at the top of the mixing sections.

For an improved version, shown in Fig. 4.24, the impeller is a volute-vaned pump which recirculates an emulsion of the mixed phases within the mixing section [Dl]. Interface control weirs provide flexibility for adjusting impeller speed and mixing intensity without upsetting the net interstage flow rate. Solvent from the previous stage flows into the vortex above the impeller and also into the interface weir section, resulting in pumping action to establish hydraulic gradients for the interstage flow of both solvent and aqueous streams.

Properties of two improved pump-mix mixer-settlers, one designed for a total interstage flow of 7.6 liters/min and the other for 380 liters/min, are summarized in Table 4.13 [Dl]. The holdup times per contactor are 1 and 2 min, respectively. The small unit was designed for reprocessing highly enriched uranium fuel and, by limiting its height to 7.6 cm, is critically safe up to Ms U concentrations of 400 g/liter, provided it is not located near any dense material that can reflect neutrons. The large unit is suitable only for process solutions of low fissile enrichment. Aqueous-to-solvent flow ratios of 0.1 to 1.5 can be attained without excessive

entrainment in the interstage flow streams. Stage efficiencies, i. e., the actual interphase transfer per mixer-settler stage relative to that predicted for an equilibrium stage, of greater than 80 percent have been obtained in the TBP extraction of uranium and plutonium [Dl].

The pump-mix mixer-settler is readily scaled over a wide range of throughputs, and because individual stages can be relied on to perform at high and reproducible efficiency, there is less risk in designing a production-scale separation plant than with some of the other types of solvent extraction contactors. A production plant can be designed with assurance on the basis of single-stage equilibrium data, data from a small-scale mixer-settler cascade, and hydraulic tests on a small section of a full-scale mixer-settler cascade. The horizontal arrangement of a mixer-settler cascade permits interruption of steady-state operation and shutdown for several hours without losing the concentration gradient of the cascade, so that the cascade can be restarted relatively easily.

Mixer-settler contactors of much larger scale are used in the solvent extraction operations associated with production of natural uranium (cf. Chap. 5), wherein nuclear criticality is not

Source-. M. W. Davis and A. S. Jennings, “Equipment for Processing by Solvent Extraction,” in Chemical Processing of Reactor Fuels,

J. F. Flagg (ed.), Academic, New York, 1961, by permission.

an issue. In the Kerr-McGee uranium extraction plant at Shiprock, New Mexico, where uranium-bearing leach liquor is contacted with alkyl phosphate in kerosene, there are four stages of mixer-settlers [Т2]. Each stage consists of a wood-stave settling tank 4.9 m in diameter and 2.1 m high in which a 1.2-m-diameter stainless steel mixing vessel with a 0.46-m-diameter turbine is placed. The aqueous flow of 6.3 liter/s is contacted with 1.3 liter/s of organic. The latter is pumped from one stage to the next by air-lift pumps. Interstage aqueous flow is by gravity, with elevation differences of 0.3 m between successive stages. The estimated holdup time per stage is 50 min.

The mixer-settler used at the Vitro uranium recovery operation near Salt Lake City, Utah, is shown schematically in Fig. 4.25. The contactor is a rubber-lined tank 6.1 m in diameter and

2.4 m straight height, capable of contacting 28 liter/s of aqueous solution [Т2]. Organic and aqueous streams from adjacent stages in the cascade are introduced directly into a turbo-mixer, with a 46-cm-diameter impeller, mounted at the top of the tank. The mixer phases emerge into the tank, which acts as a large settling chamber. For an aqueous-to-organic flow ratio of 6, the holdup time per stage is estimated to be 46 min.

Processes for recovering uranium from acid leach liquors used in the United States include solvent extraction with organic amines, solvent extraction with organophosphorus compounds, and anion exchange. Amine extraction in the so-called Amex process is described in this section with specific reference to the Kerr-McGee тШ. Solvent extraction with organophosphorus compounds in the so-called Dapex process was used in several U. S. mills, but is being phased out. It will be discussed briefly at the end of this section. Uranium recovery by anion exchange is to be discussed in Sec. 8.7.

Three papers from Oak Ridge National Laboratory provide a comprehensive summary of developments in solvent extraction of uranium from leach liquors. Coleman et al. [C2] describe studies of a number of possible amine extractants. Blake et al. [B4] describe work with organophosphorus compounds. Brown et al. [B8] describe processes based on both types of extractants.

Amine extraction. As explained in Sec. 5.4, long-chain organic amines act as liquid anion — exchange media for the uranyl sulfate complex anion through a reversible reaction such as

2(R3NH)iS04(o) + U02(S04)34-(®7) * (R3NH)4 U02(S04)3(o) + 2S04J-(e?)

The first reported use of high-molecular-weight amines to extract anions from aqueous solution was by the British workers Smith and Page [S3], who in 1948 used this method to separate strong acids from weak and suggested its use for recovering metals that form anionic complexes from aqueous solutions. Starting in 1953, Oak Ridge workers investigated a number of amines capable of extracting uranium as an anionic complex. Amines acting as liquid anion exchangers were found to be more selective in extracting uranium than organophosphorus compounds, which act as liquid cation-exchangers, because fewer of the metals associated with uranium in leach liquors form extractable anions than form extractable cations.

Of the numerous amines investigated [C2, B8], the one now used industrially in the Amex process is a mixture of straight-chain, saturated trioctyl and tridecyl amines, sold by General Mills Chemicals, Inc., under the trade name Alamine-336 and by Archer Daniels Midland Company under the name Adogen-364. Choice of this amine mixture has resulted partly from its commercial availability at an acceptable price of around $1.00/lb and partly from its having the necessary physical properties of good chemical stability, low aqueous-solubility, high uranium distribution coefficient, and good selectivity for uranium.

To use this amine, it is dissolved in a high-boiling kerosene diluent, at a concentration of around 3 v/o, approximately 0.1 M. To this solution is added around 3 v/o (0.2 M) of a long-chain alcohol, such as isodecanol, to increase solubility in the kerosene diluent of the sulfate and acid sulfate salts of the amine and its complexes with compounds of uranium and molybdenum.

Extraction equilibria with amines. Figure 5.7 shows the distribution of uranium between aqueous and organic phases observed [C2] at conditions similar to those used in the Amex process. Reported distribution coefficients range from nearly 200 in very dilute solutions to 90 to 130 at organic uranium concentrations around 0.01 M, with still lower values at higher concentrations. The leveling off of organic uranium concentration as aqueous concentration increases is attributed to approaching saturation of the amine with uranium, which would occur at 0.025 M for a complex of four amine molecules per uranium atom, as in the foregoing reaction. Physicochemical studies [C2] suggest that the complex actually contains around 4.7 amine molecules per uranium atom.

Distribution coefficients for uranium and other metals in trioctylamine are compared in Table 5.20. Although the conditions are not exactly those used in the Amex process, they indicate that the only element normally present in leach liquors that extracts readily with uranium is molybdenum. Ferric iron, which is always present in leach liquors and extracts in the Dapex process, is not extracted in the Amex process. Vanadium, if pentavalent, can be extracted by raising the pH from 1 to 2.

Figure S.7 Uranium extraction by 0.1 M tri-n-octylamine in 98 percent kerosene-2 percent tridecanol. Aqueous phase: pH = 1; 0.5 M S042 "

The uranium distribution coefficient decreases with increasing S042′ concentration, as shown in Fig. 5.8, owing to reversal of the preceding equilibrium. This permits stripping uranium from the amine by aqueous sulfate solution, as practiced at the Exxon mill, Table 5.19. The Kerr-McGee mill strips with 1.5 M Nad solution by the reaction

(R3 NH)4 U02 (S04 )3 (o) + 4NaCl(a?) ^ 4R3NHCl(o) + 2Na2S04(a?) + U02S04(aq)

Because the chloride salt of the amine is more stable than the sulfate salt or the uranyl sulfate complex, quite high aqueous uranium concentrations can be obtained with chloride stripping. Molybdenum is not stripped by sodium chloride and, if present, must be stripped by other means to prevent precipitation when its solubility of around 0.03 g/liter is exceeded.

Amine extraction in the Kerr-McGee mill. As a practical example of the use of organic amines to extract uranium from leach liquors, a description will be given of the solvent extraction section of the Kerr-McGee uranium mill, whose leaching section was described in Sec. 8.5 of this chapter [М3, Н4]. The solvent extraction plant consists of two similar circuits; process conditions approximating those of one circuit are shown in Fig. 5.9.

Leach liquor containing about 1 g U308/liter at a pH around 1.0 is fed at the rate of 3800 liters/min to the first of four mixer-settler states in series, where the uranium is extracted by a solution containing 3 v/o Alamine-336 (mixed и-trioctyl — and n-tridecylamines) and 3 v/o isodecanol in a high-boiling kerosene diluent. These four stages reduce the uranium content of the aqueous stream from 1 g U3Og/liter to around 0.001, while increasing that of the solvent from 0.002 g/liter to 3.33.

Table S.20 Distribution coefficients between aqueous sulfate solution and triisooctyl — aminet

Figure 5.8 Variation of uranium dis­tribution coefficient in 0.1 M tri-n — octylamine with aqueous sulfate con­centration. pH = 1; 0.01 mol uranium/ liter in solvent.

Each stage consists of a central steel mixer tank 8 ft (2.5 m) in diameter and 9.5 ft (2.9 m) deep set in the center of a wooden settler tank 40 ft (12 m) in diameter. Mixing is by a turbine-type impeller. Mixed aqueous and solvent phases from the central tank flow through holes in its lateral wall to the settler annulus, where the phases separate. Aqueous phase leaves through holes at the outside of the settler in the bottom and flows down to the next stage, which is set 1 ft (03 m) lower. Solvent phase is collected in a circular launder surrounding the top of the outside of the settler and is pumped up to the preceding stage. A portion of the solvent phase is recycled from the settler to the mixer to permit the latter to operate with solvent phase continuous, a condition that reduces solvent losses by entrainment in aqueous effluent.

Uranium in rich solvent leaving the extraction section is transferred to the uranium product solution by back extraction into 1.5 M sodium chloride solution flowing at the rate of 114 liters/min in four uranium-stripping mixer-settler stages. Each of these consists of a separate wood mixer tank 8 ft (2.4 m) in diameter by 9 ft (2.7 m) high and a wood settler tank 22 ft (6.7 m) in diameter by 8 ft (2.4 m) high. In this section, solvent phase flows up by gravity and aqueous phase is pumped at such a rate as to control the interphase level in each settler. Solvent recycle is unnecessary, because solvent is the continuous phase, owing to its flow rate being higher than the aqueous.

Leach liquor feed,
from Fig. 5.6
3800 .(/min
I g U308/i

Ftoffinate,
to wash failings,
Fig. 5.6

0.001 g U308/i

Figure 5.9 Amex process for recovering uranium from leach liquor. Conditions approximately those of one circuit of Kerr-McGee mill.

Product solutions leaving the two solvent extraction circuits at a concentration around 30 g U308/liter are combined and flow through four stirred precipitation tanks 8 ft (2.4 m) in diameter by 12 ft (3.7 m) high in series. Steam is added to the first tank to heat its contents to 60°C. A mixture of two to four volumes of air and one volume of ammonia is added to the last three tanks to raise the pH to 7.0. This precipitates uranium as mixed sodium and ammonium diuranate.

The diuranate precipitate is separated from the mixed NaCl and Na2 S04 salt solution by a system of thickeners and filters. Filter cake from the first filter is washed with water, reslurried with water, filtered a second time, and washed again to reduce its content of NaCl and Na2 S04. When it is necessary to reduce the amount of sodium diuranate, a third stage of filtration is used, and the filter cake is reslurried with ammonium sulfate instead of water to replace most of the sodium with ammonium ion. Washed filter cake is dried by heating to 160 to 180°C.

Salt solution leaving the filtration system contains about 0.01 g U308 and 15 to 30 g СГ/liter. Most of this is recycled to make up part of the stripping solution, but some is bled to tailings to keep sulfate ion from building up.

Molybdenum is not stripped from the amine solvent by sodium chloride. If not kept below around 0.03 g Mo/liter, it precipitates as a sludge and interferes with uranium extraction. To control molybdenum concentration, a portion of the solvent leaving the uranium stripping section is contacted in a single mixer-settler with an aqueous solution of Na2C03 and NH4OH. This converts the molybdenum to sodium molybdate, Na2Mo04, and transfers it to the aqueous phase, from which molybdenum is recovered as a by-product.

The solvent makeup requirement reported by Hazen [H4] was only 0.21 volumes per 1000 volumes leach liquor treated.

Solvent extraction of uranium with organophosphorus compounds. The first reported use of organophosphorus compounds for solvent extraction of uranium from minerals was recovery of uranium from commercial phosphoric acid using as extractant the reaction product of phosphorus pentoxide and octyl alcohol [L2]. This led to research on many organophosphorus compounds for extraction of uranium from sulfuric acid uranium leach liquors by Dow Chemical Company and Oak Ridge National Laboratory, among others. The numerous compounds investigated have been described by Merritt [М3], Blake et al. [B4], and Brown et al. [B8]. The compound selected for use in three U. S. uranium mills [М3] in the late 1960s was di(2-ethylhexyl) phosphoric acid (EHPA) dissolved in kerosene, in the Dapex process, so-named by its developers at Oak Ridge National Laboratory.

Because of the long hydrocarbon chain, EHPA and its salts are insoluble in water, but are soluble in hydrocarbons such as kerosene. The reaction by which EHPA reacts with the uranyl cation in the aqueous phase and transfers it to the organic phase may be represented by

О О

II, H „

2HOP(OR)2(o) + U022+(®7) — U02 [OP(OR)2 ] 2 (о) + 2Haq)

although the actual reaction is more complex [B4]. Thus, EHPA acts as a liquid cation — exchanger.

At the conditions typical of the Dapex process (0.1 M EHPA, 0.5 M S042′, pH = 1) distribution coefficients at very low concentration are [B4]

U022+ 200

Fe3+ 135

Al3+ 0.03

Th4+ 20,000

V4* (0.01 A0 1,000

To maintain solvent capacity for uranium and to prevent contamination of extracted uranium by iron, it is necessary to reduce iron to the unextractable ferrous condition before solvent extraction. This is done by contacting the leach liquor with scrap iron, S02, or sodium sulfide. Because the iron content of leach liquor is high, reduction is costly, and the Amex process, in which ferric iron does not extract, is preferred for sulfuric acid leach liquors. The high distribution coefficient of other polyvalent cations such as Th4+ and V4+ in EHPA makes the Dapex process less selective for uranium than the Amex process.

In the Dapex process, uranium in the organic phase is usually stripped by contact with an aqueous solution of sodium carbonate, through the reaction

О О

II II

U02 [OP(OR)2] 2(o) + 3Na2C03(a<7) ^ 2NaOP(OR2)(o) + Na4U02(C03)3(aq)

The sodium uranyl carbonate is very soluble in the aqueous phase, but the sodium salt of EHPA has only limited solubility in the organic phase and tends to form a third liquid phase containing the salt, some diluent, and water. To prevent this, the organic phase is also made 0.1 M in TBP, in which the sodium salt of EHPA is soluble.

Although the Dapex process is no longer being widely used to extract uranium from sulfuric acid leach liquors, organic phosphoric acids are favored for extracting by-product uranium from commercial phosphoric acid. Organic amines are impractical for this application because they are too fully saturated by the strong acid.