Engineering, Construction, and Operations in Space IV
American Society of Civil Engineers, pp. 1157-1166, 1994


Carlton C. Allen, Gary G. Bond, and David S. McKay


Research on oxygen production from lunar rock and soil has made significant progress in the past two years. An extensive series of experiments on natural and synthetic analogs has concentrated on high-temperature reduction by hydrogen gas. The first oxygen production experiments using actual lunar basalt, mare soil, and volcanic glass have been completed. These experiments support predictions of oxygen yield from lunar materials, based on chemical and mineralogical composition. Large samples of lunar soil simulant have been tested, and show the feasibility of scaling up oxygen production to the size of a lunar demonstration plant.


Reaction of geologic materials in a reducing atmosphere at elevated temperatures converts FeO to iron metal and releases oxygen. This is among the oldest and most technologically mature of the numerous lunar oxygen production methods which have been proposed (Taylor and Carrier, 1992). Our group at the NASA Johnson Space Center has studied the release of oxygen by high-temperature reduction using hydrogen gas. We have concentrated on three types of lunar materials: basalt, mare soil, and volcanic glass.

We have conducted over 100 experiments with natural and synthetic lunar soil analogs. These tests have identified the contributions of all major glass and mineral phases to oxygen yield. This work has recently been verified by experiments utilizing actual lunar materials. We have been able to predict oxygen yields from these materials based on chemical and mineralogical composition. In addition, we have shown that oxygen production can be scaled up to the projected size of a lunar demonstration plant.

Oxygen Production from Terrestrial Analog Materials

Experimental Technique

Powdered samples of rock, minerals and glass were reacted in flowing hydrogen in a microbalance furnace. Sample weight was monitored throughout each experiment (Allen et al., 1993). Samples weighing 0.2 - 1 g were reacted at 900 - 1100°C. Hydrogen gas flowed past the samples at a rate of 122 cm3 / min for periods of 2 - 3 hours.

Surface and internal textures were examined by scanning electron microscopy (SEM) and qualitative atomic compositions were derived by energy-dispersive x-ray spectrometry (EDS). Minerals were identified using x-ray diffraction (XRD). The oxidation states and relative abundances of iron metal and oxides were determined by iron Mossbauer spectroscopy (FeMS).


The "baseline" process for oxygen production involves reduction of ilmenite (FeTiO3), a common mineral in lunar and terrestrial volcanic rocks. The reaction, using hydrogen as a reducing agent, is:

FeTiO3 + H2 --> Fe + TiO2 + H2O

Experiments using terrestrial ilmenite have demonstrated that it reacts completely at temperatures above 800°C in flowing hydrogen (Gibson et al., 1990). Our experiments showed that reduction of the FeO is complete in approximately 30 minutes at 1100°C (Allen et al., 1992a). In addition, TiO2 is partially reduced to one of a series of suboxides with the general formula TinO2n-1. In three hour experiments at 1100°C the dominant suboxide product was identified as Ti4O7.

Limited reduction of iron oxide in other common minerals has also been demonstrated (Massieon, 1992; Allen et al., 1993). Olivine, a common phase in many lunar samples, is partially reduced at temperatures around 1100°C. Smaller degrees of reduction have been documented in pyroxene, a major mineral in most lunar basalts and mare soils.


Reduction experiments utilizing terrestrial basalt as a lunar analog have been conducted for over a decade. The results are summarized by Briggs and Sacco (1991). Our experiments have concentrated on two lunar analogs, the crystalline basalt MLS-1 (Weiblen et al., 1990) and the basaltic ash JSC-1 (McKay et al., 1994). In tests with MLS-1 FeO in ilmenite was the most completely reduced mineral, followed by olivine and pyroxene. In JSC-1 iron oxide in the glass phase was extensively reduced.


We have carried out a series of hydrogen reduction experiments on synthetic glasses which approximate the major element compositions of lunar soils (McKay et al., 1991; Allen et al., 1992a; b). Previous work on a lunar analog glass was reported by Pearce et al. (1976). Our tests have demonstrated the key advantage of glass over crystalline basalt as a feedstock for oxygen production. Basalt contains FeO in pyroxene, olivine, and ilmenite. While FeO in ilmenite and olivine reacts readily with hydrogen to release oxygen, the FeO in pyroxene reacts much less rapidly and completely. In our experiments with a glass starting material, a large percentage of the total FeO reacted to yield oxygen and iron metal before devitrification could capture the remaining FeO in the less reactive pyroxene.

Our reduction experiments with one synthetic glass powder (< 74 um) over the temperature range 1000 - 1100°C demonstrate that oxygen yield, as measured by weight loss, increases with temperature (Allen et al., 1992 a; b). Melting of the glass between 1100 - 1125°C greatly reduces the surface area and decreases the reaction efficiency.

We have also produced a range of synthetic glasses with the approximate major element compositions of several Apollo soil and volcanic glass samples (Allen et al., 1992a). These glasses have been reduced in a series of replicate experiments at 1100°C. Oxygen yield, as measured by weight loss, increases as a function of FeO content.

Oxygen Production from Lunar Materials


The first oxygen production experiments to use actual lunar material were recently carried out by Carbotek (Knudsen et al., 1992; Ness et al., 1992; Gibson et al., 1994). The experiments reacted lunar basalt 70035, crushed to <500 µm, with flowing hydrogen at 900 - 1050°C. We characterized the reacted samples by SEM, XRD, and FeMS (Allen et al., 1992c; Gibson et al., 1993a; b). Ilmenite reacted completely to iron metal plus one of several titanium suboxides. Lesser degrees of reaction in olivine and pyroxene were also documented.

Mare Soil

The most likely feedstock for a lunar oxygen plant is the local soil. Brecher et al. (1975) reacted mare soil 74241 with hydrogen in a closed capsule at 800°C. They reported a significant increase in iron metal content at the expense of ilmenite.

We have reduced mare soil 75061 at temperatures of 900 - 1050°C. This sample contains 18.02 wt.% FeO and 10.38 wt.% TiO2 (Wolfe et al., 1981), and includes abundant ilmenite grains. Ilmenite is the most reactive phase, with complete reduction of FeO in the ilmenite to iron metal at all temperatures. Electron micrographs (Figure 1) show a distinctive grain surface texture, composed of 1 - 5 µm metal belbs on a substrate of titanium oxide. Smaller degrees of reduction are indicated by iron metal blebs on the surfaces of olivine and pyroxene grains.

Figure 1

Figure 1. Ilmenite from lunar soil 75061, reduced at 1050°C, 3 hrs. Iron metal blebs cover titanium oxide. Frame width = 120 µm.

Volcanic Glass

The optimum feedstock for a lunar oxygen production process may be volcanic glass (Hawke et al., 1990; McKay et al., 1991). At least 25 distinct glass compositions exist in the Apollo sample collection (Delano, 1986). The iron- and titanium-rich species, as represented by Apollo 17 orange glass 74220 (22.0 wt.% FeO, 8.8 wt. % TiO2), promise particularly high oxygen yields. The orange glass deposit is uniformly fine grained and friable, offering a feedstock which reacts rapidly and can be used with little or no processing prior to reduction.

Brecher et al. (1975) reduced a sample of Apollo 17 orange glass with hydrogen at 800°C in a closed reaction vessel. Our reduction experiments on this same material were run at temperatures of 900 - 1100°C. In all cases significant portions of the iron oxide in the glass were converted to micrometer-scale iron metal blebs (Figure 2), with concomitant release of oxygen. At the same time, the glass devitrified to a fine-grained mixture of pyroxene and plagioclase crystals.

Figure 2

Figure 2. Glass sphere from orange glass 74220, reduced at 1100°C, 3 hours. Iron metal blebs coat devitrified glass. Frame width = 140 µm.

Prediction of Lunar Oxygen Yields

Analysis of reduced lunar basalt shows that FeO in ilmenite is completely reduced in the temperature range 900 - 1050°C, but reduction of other minerals and metal oxides is negligible (Gibson et al., 1993a ; b). Thus, the oxygen yield is independent of temperature but strongly dependent on the ilmenite content. Since the ilmenite abundance in lunar basalt is in almost all cases controlled by the availability of TiO2, the percentage of this oxide determines the rock's potential for oxygen production. If only ilmenite reacts, the predicted maximum oxygen yield from hydrogen reduction of lunar basalt equals 20% of the rock's TiO2 abundance. Further reduction of TiO2 to Ti4O7 releases oxygen equivalent to an additional 5% of initial TiO2 (Allen et al., 1992b). Thus, the predicted oxygen yield from a lunar basalt heated in hydrogen at 900 - 1050°C is 0.25 times the TiO2 abundance.

This prediction was tested against published oxygen yields from experiments on lunar basalt 70035 (Gibson et al, 1993b). This sample contains 12.97 wt.% TiO2 (Wolfe et al., 1981). The amount of water released during three experiments at 900, 1000 and 1050°C is equivalent to oxygen yields of 3.32, 3.29 and 2.93 wt.%. These values closely match the predicted yield of 3.24 wt.%.

The same prediction holds for high-titanium lunar soils such as 75061. Figure 3 shows our oxygen yield data for this soil. As in the case of lunar basalt, the yield is essentially independent of temperature, and can be predicted from the TiO2 abundance alone.

Figure 3

Figure 3. Oxygen yield (weight loss) from reduction of mare soil 75061 for 3 hours. Predicted yield equals 0.25 times the TiO2 abundance of 10.38 wt.%.

Figure 4 shows our oxygen yield data as a function of temperature for lunar volcanic glass 74220. The 800°C point was derived from the work of Brecher et al. (1975). The predicted value corresponds to the oxygen released by the complete reduction of iron oxide (0.22 x FeO wt.%).

Figure 4

Figure 4. Oxygen yield from reduction of orange glass 74220 for 3 hours. Predicted yield equals 0.22 times the FeO abundance of 22.0 wt.%.

These data, the first on oxygen production from actual lunar soil and volcanic glass, can support predictions of oxygen yields from many sites on the Moon. Cooper (1994) uses the oxygen yield from orange glass in an assessment of the resource potential of the Sulpicius Gallus region.

Large Scale Reduction of Lunar Simulants

Our work on oxygen production has also included the construction of a testbed in which large (100 - 1000 g) samples of lunar simulants can be reduced. This testbed allows us to scale up from the < 1 g samples of the microbalance experiments to the kilogram scale of a lunar demonstration plant.

The testbed was built around a Lindberg horizontal tube furnace equipped with a 6.25 cm (inside diameter) stainless steel retort. Powdered lunar simulant was loaded into an alumina "boat" and the furnace was sealed and evacuated. Hydrogen flowed past the sample as the furnace was heated and held at the operating temperature, and the sample was cooled in flowing helium. Water and other volatiles were collected from the offgas by a condenser. Solid samples were weighed before and after reduction to quantify volatile loss and oxygen evolution. These samples were then analyzed by SEM and XRD.

Our experiments utilized 125 g and 1000 g samples of lunar simulant JSC-1. Tests were run at 900 - 1100°C for periods of two hours, with hydrogen flow rates of 1250 - 6500 cm3 / min. At 1050°C and below small degrees of sintering occurred, while at 1100°C the sample partially melted. All experiments showed initial losses of volatiles, mostly water with minor chlorine and sulfur. This volatile evolution approximated the 0.71 wt.% loss on ignition (LOI) reported for JSC-1 (McKay et al., 1994). This amount was subtracted from weight losses measured in all of our experiments.

X-ray diffraction spectra of the solid reaction products from these experiments each contain a prominent peak corresponding to iron metal.

SEM analysis of reacted material from the 1050°C experiments shows abundant iron metal blebs, as large as 2.5 µm across, decorating the surfaces of glass grains. Blebs smaller than 0.5 µm sparsely cover the surfaces of olivine crystals. The metal apparently forms by reduction of FeO in the interiors of these grains and migrates to the surfaces to produce blebs. Samples reduced at 900 and 1000°C exhibit iron metal within the glass grains, but few blebs on their surfaces.

Weight loss due to oxygen release is correlated with reaction temperature. Figure 5 shows the weight losses (after LOI) for 125 g samples of JSC-1 reduced at 900 - 1050°C. The temperature dependence, similar to that seen for lunar volcanic glass (Figure 4), indicates that glass is the major reduced species in JSC-1.

Figure 5

Figure 5. Oxygen yield from reduction of JSC-1 for 2 hours.

JSC-1 contains 7.35 wt% FeO and 3.44 wt% Fe2O3 (McKay et al., 1994). If all of this oxygen were released the weight loss would be 2.67%. A 1 g sample reduced at 1050°C for two hours in the microbalance furnace lost 1.92 wt.%. Figure 5 shows a loss of 1.70 wt.% from a 125 g sample heated to 1050°C in the testbed. A 1000 g sample lost 1.28 wt.%.

The decline in oxygen yield with increasing sample size may be due to any of several factors. The heating time or testbed furnace configuration may not be sufficient to bring the entire 1000 g sample to 1050°C. Hydrogen and water may not be able to diffuse through all portions of the large samples. Finally, the rate of hydrogen flow may not be high enough to efficiently reduce a kilogram of JSC-1. We are continuing our experiments to define which scaleup parameters will most affect lunar oxygen production.


High-temperature hydrogen reduction experiments on terrestrial analogs to lunar materials show that ilmenite and glass are the most reactive phases, followed by olivine and pyroxene. Tests on lunar basalt, soil, and volcanic glass confirm these results. The yield from basalt and mare soil is a function of composition, while the yield from glass is sensitive to both composition and temperature. Maximum oxygen yields from high-titanium soil and iron-rich glass are 3.0 and 5.4 wt.%, respectively. Tests on < 1 g samples are being scaled up to masses of 1000 g, appropriate to the size of a lunar demonstration plant.


We gratefully acknowledge R.V. Morris, H.V. Lauer, Jr., A.J.G. Jurewicz, C. Romanek, and J.L. Winkler for experimental and analytical support, and D.C. Lynch and L.A. Taylor for helpful reviews. Experiments were performed at the NASA Johnson Space Center Igneous Petrology Laboratory with support from the Center Director's Discretionary Fund.


Allen C.C., McKay D.S., and Morris R.V. (1992a) Lunar oxygen - the reduction of glass by hydrogen. Engineering, Construction, and Operations in Space III, pp. 629-640. American Society of Civil Engineers, New York, NY.

Allen C.C., McKay D.S., and Morris R.V. (1992b) Hydrogen reduction of lunar simulant glass. Lunar and Planetary Science XXIII, pp. 21-22. Lunar and Planetary Institute, Houston, TX.

Allen C.C., Gibson M.A., Knudsen C.W., Kanamori H., Morris R.V., Keller L.P., and McKay D.S. (1992c) Lunar resources -- oxygen from rocks and soil. Workshop on geology of the Apollo 17 landing site, LPI Technical Report Number 92-09, Part 1, pp. 1-2. Lunar and Planetary Institute, Houston, TX.

Allen C.C., Morris R.V., Lauer H.V. Jr., and McKay D.S. (1993) Microscopic iron metal on glass and minerals -- a tool for studying regolith maturity. Icarus, 104, pp. 291-300.

Brecher A., Menke W.H., Adams J.B., and Gaffey, M.J. (1975) The effects of heating and subsolidus reaction on lunar materials: an analysis by magnetic methods, optical, Mossbauer, and x-ray diffraction spectroscopy. Proc. Lunar Sci. Conf. 6th, pp. 3091-3109.

Briggs R.A. and Sacco A. Jr. (1991) Hydrogen reduction mechanisms of ilmenite between 823 and 1353 K. J. Materials Res., 6, pp. 574-584.

Cooper B. (1994) Reservoir estimates for the Sulpicius Gallus region. This volume.

Delano J. (1986) Pristine lunar glasses: criteria, data, and implications. Proc. Lunar Planet. Sci. Conf. 16th, in J. Geophys. Res., 91, D201-D213.

Gibson M.A., Knudsen C.W., and Roeger A. III (1990) Development of the Carbotek process for lunar oxygen production. Engineering, Construction and Operation in Space II, pp. 752-762, American Society of Civil Engineers, New York, NY.

Gibson M.A., Knudsen C.W., Brueneman D.J., Kanamori H., Ness R.O., Sharp L.L., Brekke D.W., Allen C.C., Morris R.V., Keller L.P., and McKay D.S. (1993a) First oxygen from lunar basalt. Lunar and Planetary Science XXIV, pp. 531-532. Lunar and Planetary Institute, Houston, TX.

Gibson M.A., Knudsen C.W., Brueneman D.J., Allen C.C., Kanamori H., and McKay D.S. (1993b) Reduction of lunar basalt 70035 - oxygen yield and reaction product analysis. Submitted to J. Geophys. Res.

Gibson M.A., Knudsen C.W., Brueneman D.J., and Kanamori H. (1994) Kinetic interpretations of first reactivity experiments on lunar soil samples. This volume.

Hawke B.R., Coombs C.R., and Clark B. (1990) Ilmenite-rich pyroclastic deposits: an ideal lunar resource. Proc. Lunar Planet. Sci. Conf. 20th, pp. 249-258.

Knudsen C.W., Gibson M.A., Brueneman D.J., Suzuki S., Yoshida T., and Kanamori H. (1992) Recent developments of the Carbotek process for production of lunar oxygen. Engineering, Construction and Operation in Space III, pp. 597-605, American Society of Civil Engineers, New York, NY.

Massieon C.C., (1992) Reduction of iron-bearing lunar minerals for the production of oxygen (MS Thesis). Univ. of Arizona, Tucson, AZ, 106 pp.

McKay D.S., Morris R.V., and Jurewicz A.J. (1991) Reduction of simulated lunar glass by carbon and hydrogen and its implications for lunar base oxygen production. Lunar and Planetary Science XXII, pp. 881-882, Lunar and Planetary Institute, Houston, TX.

McKay D.S., Carter J.L., Boles W.W., Allen C.C., and Allton J.H. (1994) JSC-1: a new lunar soil simulant. This volume.

Ness R.O. Jr., Sharp L., Brekke D.W., Knudsen C.W., and Gibson M.A. (1992) Hydrogen reduction of lunar soil and simulants. Engineering, Construction and Operation in Space III, pp. 617-628, American Society of Civil Engineers, New York, NY.

Pearce G.W., Hoye G.S., Strangway D.W., Walker B.M., and Taylor L.A. (1976) Some complexities in the determination of lunar paleointensities. Proc. Lunar Sci. Conf. 7th, pp. 3271-3297.

Taylor L.A. and Carrier W.D. III (1992) The feasibility of processes for the production of oxygen on the Moon. Engineering, Construction and Operation in Space III pp. 752-762, American Society of Civil Engineers, New York, NY.

Weiblen P.W., Murawa M.J., and Reid K.J. (1990) Preparation of simulants for lunar surface materials. Engineering, Construction and Operation in Space II, pp. 428-435, American Society of Civil Engineers, New York, NY.

Wolfe E. A., Bailey N.G., Lucchitta B.K., Muehlberger W.R., Scott D.S., Sutton R.L., and Wilshire H.G. (1981) The geologic investigation of the Taurus-Littrow valley: Apollo 17 landing site. Professional Paper 1080, Geological Survey, Washington, DC, 280 pp.

Lockheed Engineering & Sciences Co., Houston, TX

Space Engineering Research Center, University of Arizona, Tucson, AZ

NASA Johnson Space Center, Houston, TX