During the last few years of employment at GE's lighting headquarters at Nela Park in E. Cleveland, Ohio, the writer's interest in the history of lighting, as well as the people involved in its development, increased greatly. A brief description of some of the persons involved with the fluorescent lamp has already been presented on this website (see the last entry of Section 16, Biographical Sketches). A brief biographical sketch of Dick Thayer was presented there.
Dick Thayer started to work at Nela Park in the year 1928. Forty-one years later, in 1969, he retired from the position of Manager of Fluorescent Engineering (Obituary, "Richard N. Thayer, Led Fluorescent Lamp Creation," The Plain Dealer, Cleveland, Mar 29, 1992, pg 10B). While in retirement, during 1988-89, Thayer wrote a narrative of the development of the fluorescent lamp as he witnessed it. He kept the writing to himself and, apparently, did not make it known to those in the industry. In September of 1991 this writer decided to visit Thayer so that his personal recollections about the development of the fluorescent lamp could be recorded. During the visit he left the room and returned a short while later with a rough draft of the lamp development history he had written; the writing was a complete surprise to me. I immediately suggested that he allow me to take the writing to Nela Park for publication as an internal report. He agreed to my suggestion and the writing was typed by a secretary. I withheld his drawings just long enough so that they could be put in a form suitable for a report. The report was issued on October 25, 1991.
It was recently decided by this writer that Thayer's report would be of interest to the public because such a narrative by a person involved in the development of a important product is not all that common. The writer therefore contacted the Legal Department at Nela Park to see if approval for publication on this website would be possible.
The writer is grateful to Timothy B. Gurin, Senior Counsel at Nela Park, and the General Electric Company, for granting permission to publish Dick Thayer's report on this website.
The purpose of presenting this material is to add to our lamp history knowledge base and it is not presented for personal or monetary gain by anyone, including this writer. In the event someone, or some organization, desires use of any of this material, permission to use it should be obtained from the Legal Department of the Lamp Business Group of the General Electric Company at Nela Park, East Cleveland, Ohio.
I am grateful that Dick Thayer, one of my former managers, had the interest and inclination to write this account of the early development, within the General Electric Company, of the fluorescent lamp. It is the writer's view, based on reading early internal reports, that Dick was an exceptional experimenter with both incandescent and fluorescent lamps. He continued to contribute through his management skills. The writer hopes interested readers will enjoy his narrative.
Thanks are again extended to Timothy B. Gurin at Nela Park for his efforts in allowing this report to appear on this website.
R. N. Thayer
The First Report
But within the lifetimes of many of us, it was not always so. Our acquaintance with fluorescence was then limited to high-school science classes. There, our teacher would draw the shades to darken the room, turn on an incandescent or small mercury-vapor lamp with purple-glass filter, and shine it on selected mineral specimens. Then, to our dark-adapted eyes, we would see them glow in a variety of colors. It didn't seem possible that this darkroom phenomenon could ever be the basis for a practical light source.
This view was confirmed by the conclusions of scientists and inventors in the field. Edison, himself the author of an 1896 patent entitled "Fluorescent Lamp," wrote after experimenting to 1910 that "the project with the fluorescent lamp is hopeless." And E. L. Nichols, another experimenter in the field, wrote in 1921 that "unless someone discovers a means of making luminescent bodies that are vastly brighter than the best now known, luminescence (i. e. fluorescence) may be excluded altogether as a factor in artificial lighting." At the General Electric Vapor Lamp Co. in Hoboken, NJ (formerly Cooper-Hewitt), various attempts were made to use fluorescent materials to improve the color and /or light output of their medium-pressure tubular mercury-arc lamps, which are rich in the ultra-violet radiation most commonly used to produce fluorescence. They obtained little or no gain, and were convinced that fluorescence had no commerical usuage. In 1933, we at the Lamp Development Laboratory of the General Electric Co. in Cleveland obtained similar negative results. Our tests used a cathode-glow type of mercury-arc lamp, contained in an incandescent sized household bulb.
Imagine our surprise, then, when we received a report in August 1934 that the laboratory of the General Electric Co. Ltd. in London (no relation to U. S. GE) had made experimental green fluorescent lamps with an efficiency of 35 lumens-per-watt (LPW), three times that of household-size incandescent lamps! This report came to us from Dr. Arthur H. Compton, the Nobel prize-winning physicist and then a consultant for our laboratory, then travelling in Europe. We thought that the decimal point must be wrong, and that the correct figure was 3.5 LPW. But further European information was confirming, and in December 1934 we began U. S. development on a lamp obviously much more complicated and expensive than the incandescent. This narrative covers our three-year development until the announcement of commercial availability in the spring of 1938, and the first major public use at the New York City World's Fair in 1939.
With the great eventual success of the fluorescent lamp, it is often asked "Who invented it?" We have no simple answer. Our earliest U. S. source of information was three European lamp companies with whom we then exchanged technical information: the General Electric Co. Ltd. of London, already mentioned; the Philips Co. of Eindhoven, Netherlands; and the Osram Co. of Berlin, Germany. We don't know the division of credit or priority among them. In this matter, a later decision of the General Electric Co. (U. S.) was to purchase the mid-1920's patent of three independent German inventors, - Meyer, Spanner, and Germer. But while that patent describes the low-pressure mercury discharge lamp, it makes no mention of any fluorescent coating.
In some publications, the invention is ascribed to Edison, despite his disclaimer above. His 1896 patent bears no relation to the successful design. And John L. Enos of MIT, in 1965 testimony before a Congressional committee, attributed Becquerel, a French scientist in this field, the 1859 (!) "earliest conception of the product in substantially commercial form." So, he indicated a 79-year interval (delay) from invention to innovation. Needless to say, we disagree completely.
Our view is that the successful development resulted from a combination from the three separate areas: (1) oxide-coated hot-cathodes; (2) inside fluorescent coatings on discharge tubes; (3) a specialized design of low-pressure mercury-vapor discharge for high-efficiency resonance radiation. Details on each of these will follow.
At the Incandescent Lamp Department in Cleveland, two development groups were eventually set up. The first and principal one was in the Lamp Development Laboratory under George E. Inman as group leader. This group operated fairly independently, since the higher Laboratory managers were then principally concerned with the photoflash and sealed-beam auto headlight developments, - two other major programs under way at the same time.
About half way thru the 1935-38 development period, when eventual commercialization seemed likely, a second Cleveland group, headed by P. J. Pritchard, was set up. Their objective was to adapt, develop, and build automatic equipment to manufacture the lamp. This group had past experience with radio-tube manufacture, which has some similarities. A noteworthy volunteer to this group was T. W. Frech, then just recently retired as chief executive of the Incandescent Lamp Department, and a former president of the Peerless Motor Car Co. He was a forceful and enthusiastic experimenter with this new light source. While some of his test procedures and conclusions were amateurish, his former rank obviously procured him a respectful top-level hearing, and he made direct contributions to the final designs.
Curiously, these two Cleveland development groups operated rather independently, - more in rivalry than in cooperation. In certain design areas, process and equipment decisions made by the Pritchard group effectively settled disputed design choices, as will be detailed later. At a later development stage, Ward Harrison, a highly capable executive engineer, who was Manager of Application Engineering, was appointed co-ordinator and final decision maker.
Lumiline End-Disc and Electrode Structure
The universal early view of the fluorescent lamp's future was that its usuage would lie chiefly in the colored-lamp and decorative field. Only its very high efficiencies in this area appeared to justify its much greater complexity and cost vs. incandescent. To minimize development time and cost, Inman decided to use or modify design and equipment already developed for the commercially available Lumiline Lamp, - an 18" long, 1" diameter tube with a single disc contact at each end (Fig. 2) and a tungsten filament stretched and supported along its axis. It was designed for such uses as showcase and bathroom mirror lighting. A further objective, set early by Inman to assist public acceptance and reliability was "instant-starting, with no moving parts (i. e. no starting devices)." These goals were laudable and at first widely accepted, but their implementation encountered severe problems in certain areas, leading to design alternatives as we will detail later.
While the largest development effort took place at Cleveland, as described here, parallel efforts were proceeding elsewhere, - at Westinghouse, at Hygrade-Sylvania, and at GE Vapor Lamp Co. in the U. S., and in Europe. This article concentrates on Cleveland and on our personal participation (Fig. 3).
The first phosphors we used in early 1935 were naturally occurring minerals long known, - zinc silicate (willemite) for green, cadmium silicate for pink, and calcium tungstate for blue. Rocks of willemite were ground by hand to dust in a mortar and pestle, and then applied as a bulb coating. We were soon able to make brighter phosphors synthetically. To make zinc silicate phosphor, for example, appropriate quantities of zinc oxide, silica, and manganese oxide are thoroughly mixed dry, then fired in air for one or two hours at 1100-1200 C. Our early batches of thimble-sized crucibles have since grown to tons per day.
The search for and successful discovery of improved phosphors has continued ever since, on an international scale. This search is highly empirical; theoretical explanations and predictions of fluorescence have been reached only for simple compounds of only academic interest. In the period covered by this article, the two most important phosphors developed were magnesium tungstate (blue-white color) and zinc beryllium silicate (yellow-white color.) The first was known to us almost from the beginning; we learned of the second from our Research Laboratory, at about the middle of our development. In various formulations and mixtures, these two phosphors together can produce a wide range of white colors that dominated the later development. No toxicity hazard was then known for the beryllium compound, and many of us in development worked daily with it, with no ill effects. It has long since been replaced by a non-toxic complex halophosphate used throughout the industry.
In early development, fluorescent lamps in certain colors showed tremendous gains in efficiency and light output over their incandescent counterparts. In green, the efficiency gain was fifty times, and in blue, two hundred times. Incandescent lamps make inefficient color sources because filter coats must absorb most of their light output. In fluorescent lamps, the color is inherent in the particular phosphor. These advantages produced our early view of fluorescent lamps as most attractively a colored light source, as already mentioned. For 1938 commercialization we listed five colors, - green, blue, gold, pink, and red. Gold and red still required adding a color filter for desired color purity, because the bluish color from the discharge itself diluted the phosphor color.
From the beginning, we recognized that white lamps could be made by appropriate mixes of blue, red, and green phosphors, but efficiency gains over incandescent were far smaller than in colors, - only about three times. Also, such mixes were highly sensitive to the green-phosphor addition; too little, and the resulting color was purplish, too much and it was greenish. The two-component white phosphor mix developed later (above) proved much easier for color control because both constituents were broad-band emitters.
A further advantage of fluorescent lamps over incandescent was the ease with which a daylight color could be obtained by a suitable phosphor mixture, - an aim long sought in incandescent without success. Our 1938 introduction, then, included two white colors, - daylight, and a warmer color called 3500 degree white (matching an incandescent source of that color temperature). Acceptance of the daylight color has been disappointing. Even though it matches the color of the light coming in our windows, users have judged it too "cold" and bluish. But a white color called "cool white," with a color temperature of about 4200 K, has become by far the most popular color, amounting to over half of all fluorescent lamps sold.
Our earliest coating method was called "dusting." The ground phosphor particles were still large enough to flow freely, like sugar or salt. Next, a sponge soaked with glycerine, plus an additive of some low-melting glass compound, was run thru the tube to provide a thin, sticky layer. The tube was then held at an angle from the vertical, and rotated slowly as the phosphor flowed in from the top. A uniform, grainy coating resulted. The coated tube was then rolled thru a hot oven (lehred) to remove the glycerine and melt the additive. This method produced good lamp efficiencies, but coating adherence was weak. Jars or bumps could loosen small or large parts of the coating.
After a few months, we shifted to a paint-type coating we called "poured coatings," still the basic coating method. We had first avoided it because the brightness of phosphors is seriously reduced by the long milling required to suspend them in paint solvents. We first learned from Westinghouse that this loss could be reduced to an acceptable level by using a lacquer-type suspension with carefully controlled ball-milling. Also, the viscosity of the suspension is increased by adding nitrocellulose to it; this aids in the suspension of larger particles. The bulb was then coated as before, but with liquid pouring from the top. In later automatic equipment developments, the phosphor suspension has been either "upflushed" or "down-flushed" into the bulbs.
The coated tube must then be dried, and lehred up to nearly the softening point of the glass, to remove all traces of the solvent and nitrocellulose. Coating adherence is still excellent because of the fineness of the individual phosphor particles.
The anode is customarily a bent wire or pieces of sheet metal that serves as a current collector on the half-cycle when that electrode is positive. Occasional experiments were made on its design, but its role is only an auxiliary one. Cathode design is the most important determinant of lamp life. It is usually a coiled-coil of fine tungsten wire, much like that used in lower-wattage incandescent lamps. To obtain the desired electron emission at low energy and low temperature, it is coated with an "emission mix" of mixed barium-strontium-calcium carbonates. At later exhaust, this coated cathode must be heated to decompose the carbonates to oxides. This emission mix is gradually used up thru the life of the lamp; when it is all gone, the lamp is said to be "deactivated," and will no longer start or run.
Cathode design variations for longer lamp life, easier starting, reduced end discoloration, and lower cost have been frequent from the beginning of the development. Details are too numerous to include. But interestingly at one time in development, rival cathode designs from the Inman and Pritchard groups necessitated the intervention of a neutral physicist from GE's Research Lab., who ruled in favor of the Pritchard design.
The "cold-cathode" design used in neon tubes was once considered a rival to the standard "hot cathode" described above. Its severe limitation is an electrode energy loss (voltage drop) of 100 volts or more, vs. 15 volts for hot cathodes. See Fig. 4. Further, its top current limit is commonly 0.12 amps. Competitive lines of standardized-length cold-cathode lamps failed to obtain wide acceptance for these reasons.
Starting frequency is a factor in fluorescent lamp life, and we early set a test schedule at one start for each three hours of burning, - a cycle that has become industry standard. One attempt we made to predict life was accelerated starting tests, - on and off every few seconds. With the instant-starting of our original Lumiline design, the sputtering of emission mix at every start limited life to a few hundred starts. Later, with full cathode preheating on a switch-start circuit, lamps usually have lasted fifty to a hundred thousands starts. We obtained no good correlation with regular burning.
Life tests also show how well the lamp retains its initial light output, a factor called lumen maintenance. In early tests, this varied widely with individual phosphors. Tungstates were relatively high, and green zinc silicate was low.
On first lightup after manufacture, fluorescent lamps may be unstable for a few hours in their electrical and light characteristics. So, we initiated the practice of burning lamps for 100 hours before measuring and specifying initial ratings. This is still industry practice.
The early instant-start experimental lamp designs lasted only a few hundred hours on the standard 3-hour cycle. With the later change to starter-switches and cathode preheat, our first listed life rating of 1000 hours was easily achieved, and was soon increased to 2500 hours.
A further exhaust problem for fluorescent lamps is the evolution of large quantities of carbon dioxide when the emission mix is first heated, - a step called "decomposition" or "activation." Our early development trials of other barium compounds, to reduce or eliminate this problem were not successful, and carbonates remain standard.
This problem was particularly severe for the early Lumiline design, already described. Since it had only one contact, a metal disc, at each end, the cathode could not be heated in the common way, by a current thru it. Instead, we had to use a method we called "arc activation." After initial exhaust, we admitted some argon gas into the lamp, then initiated an arc across it from a 2000-volt source. The heat of the arc evolved carbon dioxide from the emission mix, and quenched the arc. This was exhausted, a fresh dose of argon admitted, and the process repeated. After three or four steps, decomposition was adequate, and final exhaust continued. However, this "brute force" procedure consumed much of the emission mix from bombardment and sputtering. Our efforts to redesign the disc for double contact failed.
When the Pritchard group began exhaust equipment design, they unilaterally discarded this approach, used two wires at each end to heat the cathode as they had done in radio-tube manufacture, and proceeded to make the first designs of two-contact bases and sockets, - standard ever since (Fig. 5).
The last part of exhaust includes the addition of a tiny mercury droplet (50 mg.) about the size of a pinhead, but ample for the low mercury vapor pressure required. A second addition is a filling of about 3 mm. argon; this is required to facilitate starting, and to reduce electrode drop.
An electric discharge thru a gas or vapor produces a spectrum of many lines, scattered thru the visible, ultra-violet, and infra-red. In mercury-vapor discharges, for example, the visible spectral lines are in the purple, blue, yellow, and green, resulting in its characteristically bluish-green color. The energy emitted in each spectral line is typically 1% or less of the total energy input to the discharge. But in the successful fluorescent lamp design, about fifty percent is converted into one spectral line, - the 2537A resonance-radiation line of mercury.
Relative Output and Efficiency of Light and of Mercury-Resonance
Radiation in Fluorescent Lamps, vs Bulb-Wall Temperature
Attainment of this high conversion efficiency requires several conditions. First and foremost, the mercury vapor pressure must be at or close to 10 microns (1/76,000 atmosphere). This very low pressure is reached at a bulb-wall temperature of about 40 C (105 F). See Fig. 6. This is far below the hot bulb-wall temperatures of an incandescent or high-pressure mercury vapor lamps. Departures from this optimum value cause serious losses. This is easily noted in the dim light output from fluorescent lamps exposed to cold weather outdoors.
Second, the lamp must be a positive-column design; i. e. one where the arc length is at least several times the bulb diameter. This minimizes the percentage of end losses at the electrodes. See Fig. 4. Third, current-density in the discharge must be low; for example, about 1/4 ampere in a 1" diameter bulb.
All these requirements indicate preference for a low, slim, high-voltage, low-current, low-wattage, cool-bulb design.
Our earliest experimental lamps were 3/4" diameter and 12" long. We soon shifted to the 1" diameter 18" long Lumiline bulb, as noted above. Industry practice designates this as the 18"T8; T stands for tubular shape, and 8 for the diameter in 1/8" steps. The Inman group kept this size thruout development as its standard, and it was commercialized as the 15-watt T8, at about 1/4 amp and 60 volts.
Frech proposed and pushed a rival design, the 24"T12 at 20 watts. Voltage was about the same, current higher, efficiency closely the same, and light output 40% higher. Obviously, it was a larger, more expensive design. The main advantage claimed for it was lower brightness, as it was then expected that fluorescent lamps could be used bare because of their far lower brightness than incandescent. Efforts to resolve the T8/T12 design rivalry, including a compromise on T10, failed, and both sizes were commercialized. Both diameters have survived in the market place to this day, and are still competing in new, long-length designs.
Two other designs were included in first commercialization. See Table 1. One was the 18"T12 (15wT12), a "straddle" between the two competing designs. The fourth design was a 36"T8 (30wT8), doubling the length of the 15wT8, and intended for use on 208 or 240-volt industrial circuits.
The earliest concept, described above, of a Lumiline fluorescent lamp with only a single disc at each end, and "instant starting with no starters" required a high open-circuit voltage to initiate the arc without any cathode preheat. Our testing set this voltage at 450, and Ft. Wayne designed a high-leakage reactance transformer that combined the functions of transformer and ballast (Fig. 7B). It was bulky, expensive and high-loss, - about 12 watts just to operate a 15-watt lamp.
Moreover, lamps occasionally failed to start, even at that high voltage. We observed that if a hand touched or came near the bulb, it would start. This led to the design of a "starting stripe," a conducting carbon or metallic stripe drawn along the outside of the bulb, and connected to one end. This stripe was baked on at high temperature. But occasionally scratching or bumping would break its electrical conductivity, and its effectiveness was lost. When we transferred to switch-start circuits, it was no longer needed, and was dropped.
The developing handicaps of arc activation (above) and ballast for the early Lumiline design concept were serious, but Inman continued to prefer it over the other complexities of double-contacts at each end, and switch-start circuit design. The Pritchard group, however, took this step, and it resulted in a far smaller and cheaper series-inductance ballast, called a reactor (Fig. 7C.) In essence, a reactor is a coil of wire wound around a stack of magnetic-iron sheets, and it had only about 5 watts loss. Various starter-switch designs are possible, including the manual one still used on fluorescent desk lamps. Dench of Westinghouse invented an ingenious two-contact glow-switch starter, and this combination became the eventual standard for all switch-start circuits.
Frech deplored the five-watt ballast loss, which he regarded as excessive for operating a 15 or 20-watt lamp. He designed and built a much larger and costlier ballast that consumed only 1.5 watts. His design also incorporated an integral magnetic starter that operated from the magnetic field of the ballast. As with a buzzer or old-fashioned doorbell, a moving arm made intermittent contact until the cathodes heated and the arc struck. Several thousand of these were made and offered for sale, but failed commercially because of high cost, - ten times that of the simple ballast - large size, and troubles with the delicately balanced vibrator mechanism.
Another discarded type was the resonant-circuit ballast, another effort to eliminate the need for a starter switch. In one form, an inductance and a capacitor are connected in parallel with the lamp (Fig. 7D). When the circuit is switched on, these combine with the series ballast to supply both high cathode heat, and high voltage across the lamp. When the lamp starts, the parallel circuit is shunted out, and operating current is normal. Some of these were tried commercially in the 1939 New York World's Fair. But if the lamp is deactivated or otherwise fails to start, the high starting currents and voltages continue, and overheat the ballast. The large size and cost required to provide for this caused its abandonment.
Still another ballast design of short commercial life was the combination in a single metal box of the simple ballast and a four-contact thermal switch (before the Dench invention). As with the Frech design, a weakness of this was that shipping and handling of the heavy assembly could easily jar the delicate starting switch out of adjustment. These troubles with integral switches led to the final concept of a separate, easily replaceable switch, which has been standard ever since.
An interesting and valuable feature of all switch-start circuits is that the opening of switch contacts, after cathode preheat, produces a several-hundred-volt transient, from the collapse of the ballast's magnetic field. This has proved an important aid to reliable lamp starting.
The traditional way to correct low lagging power-factor, which this is, is by the addition of capacitors. They are large and expensive. The problem was finally resolved by the development, by M. A. Edwards of GE's General Engineering Laboratory, of a two-lamp circuit called "lead-lag." See Fig. 8. Both lamps (30w or 40w) operate from the secondary of a 120/200V stepup transformer. One lamp has the normal reactor in series, and 0.5 lagging power-factor. The second lamp operates from a reactor and capacitor in series, with a 0.5 leading power-factor. The combined circuits on the primary side have a power-factor above 0.9. The added capacitor cost is minimized in this circuit. In addition, the two lamps operate sufficiently out of phase to cut in half the 60-cycle flicker of the lamps that is annoying in some uses. The industry adopted this circuit for all the longer lamps - three feet or more - which have constituted the major load for the utilities.
Magazine of Light, published by the Incandescent Lamp Department, General Electric Co., Vols. 7 & 8, 1938-39.
"The Basis for High Efficiency in Fluorescent Lamps," R. N. Thayer and B. T. Barnes, Journal of the Optical Society of America, March 1939, pp 131-134.
Fluorescent Lighting Manual, Charles L. Amick, McGraw-Hill Book Co., 1st editon, 1942.
"Economic Factors Influencing Development and Introduction of the Fluorescent Lamp," A. A. Bright, Jr. and W. R. MacLaurin, Journal of Political Economy, Vol 51, 1943, pp 429-450.
Fluorescent and Other Gaseous Discharge Lamps, W. E. Forsythe and E. Q. Adams, Murray Hill Books, 1948.
"Fluorescent Lamps, - Past, Present, and Future," G. E. Inman, General Electric Review, July 1954, pp 34-38.
Lamps for a Brighter America, P. W. Keating, McGraw-Hill Book Co., 1954.
Economic Concentration Hearings, 89th Congress (1965), Subcommittee on Patents, pp 1481-1491.
A Streak of Luck - The Life and Legend of Thomas A. Edison, Robert Conot, Seaview Books, 1979.