Fifty Years of the Gas-Filled Lamp

2 PHILIPS TECHNICAL REVIEW VOLUME 25

FIFTY YEARS OF THE GAS-FILLED LAMP

by J. C. LOKKER *). 621.326.72

The invention of the gas-filled lamp, now hal] a century ago, was one of the more important advances ~ perhaps the most important ~ in the evolution of the incandescent lamp. This 50th anniversary gives us occasion to trace once again the development of the gas-filled lamp. The share which Philips had in tliis development is also recalled in the article below. Mr. Lokker, uho wrote this article at our request and whose photograph appears here, took an active and leading part in the development of the incandescent lamp at Phiiips from its earliest days. He was the first graduate engineer to be appointed by Mr. G. L. F. Philips, joining the company in 1908. In later years he managed the department now referred to as the "Lighting Division", until his retirement in 1945.

In the year 1879, Thomas Alva Edison solved the fabriken" (Associated Lamp Manufacturers' Selling problem of how to produce light with electricity in Agency). While this brought some peace on the a reasonably practical form. The invention was commercial side, there was no easing-up of pressure demonstrated with great success at the Paris World on the production side, caused by the demand for Exhibition in 1881. This was the beginning of the new and better lamps. Although the production of carbon-filament lamp. Several firms and engineers carbon filaments was substantially improved by set to work to make similar lamps and installations, changing from zinc-chloride cellulose to collodion but since electricity networks were few and far acetate as the basic material, the carbon filaments between, the development and spread of electric nevertheless consumed too much power for a given light made slow headway at first. light output, and the lamps tended after some time G. L. F. Philips, born in 1858 and who graduated to turn black. Efforts made to find other materials at Delft in 1883 as a mechanical engineer, was so for the filament led for example to the Nernst lamp, interested in the principles of electricity, and espe- which used a slender rod of thorium-cerium oxide cially in the carbon-filament lamp, that after a few (1897), the osmium lamp (1900), tbe tantalumlamp years of gaining practical experience, and after ex- of Siemens (1904) and, in 1906, the tungsten-fila- perimenting in a primitive laboratory in his parents' ment lamp (fig.I). home at Zaltbommel, he began in 1891 to manu- Although the melting point of tungsten is not so facture carbon-filament lamps in a former buckskin high as that of carbon, its rate of evaporation at factory at Eindhoven 1). At that time other firms high temperature is much lower. This made tungsten were already producing these lamps in large quanti- better suited as a material for lamp filaments. The ties, and in the early years G.L.F. Philips had to melting point of tungsten metal was too high, how- contend with numerous difficulties. If his brother ever, for it to be melted in any material known at A. F. Philips had not come to his assistance in 1895 the time (graphite was ruled out, for chemical to organize the selling side of the business, he might reasons), and therefore a special method had to be well have had to close down production. Gradually devised for obtaining tungsten in the form of wire. the business began to prosper. Since there were For this purpose a very fine powder of tungsten was hardly any electric power stations in the Nether- mixed with an organic binder to form a paste which lands in those days, the two brothers turned their was "squirted" through fine holes in diamond dies. attention to the German market, to such good effect After pre-heating in an inert gas to remove the organ- that the Düsseldorf Gewerbeausstellung (Industrial ic binder, followed by heating to a very high tem- Exhibition) in 1902was lit entirely by Philips lamps. perature (the preparation process), filament wire The bitter competitive struggle fought with other with a bright metallic surface was obtained. The fil- manufacturers led in 1903 to the setting-up in Berlin aments were put on special mounts and sealed in of the "Verkaufsstelle Vereinigte Glühlampen- glass bulbs. The lamp so produced, which came out in about 1906, was called the "squirted"-tungsten- *) Formerly with Philips, now in retirement. filament lamp. 1) See N. A. Halbertsma, The birth of a lamp factory in 1891, Philips tech. Rev. 23, 222-236, 1961/62. This lamp was a very considerable improvement 1963/64, No. 1 HALF -WATT LAMP 3

Photo Science Museum, London Fig. 1. Six metal-filament lamps from the years 1897 to about 1937.

on the carbon-filament lamp, which therefore grad- shocks and transportation proved to be a drawback. ually disappeared from the market, although it Every possible endeavour was therefore made to continued to be used for years in places where the find a method of making stronger tungsten filaments, lamps were subject to severe vibrations. The reduc- viz, by drawing. tion of the power consumed, from 3.5 W per candle The first to succeed was the American Coolidge in the carbon-filament lamp to about 1 W in the in 1908. In his process the tungsten powder was tungsten lamp, quickly proved decisive both as pressed into thin bars, which were pre-heated to regards the potential uses of the electric lamp and make them conductive and to give sufficient co- the spread of power stations. The method of manu- herence for handling, subsequently further heated facturing the new lamps, however, more resembled in an inert gas to just below their melting point, and laboratory work than factory production, and the then machine-hammered white-hot (swaged) into fact that they were not well able to withstand thin rods (fig. 2). These operations made thematerial

Fig. 2. Ductile tungsten wire is made by machine-hammering (swaging) sintered tungsten bars to increase their density after which they are drawn to the required thickness through hard-metal or diamond dies. The photograph shows a swaging machine, with the tungsten bar being introduced manually after having first been raised to a very high tempera turc in the adjoining furnace; after some passes through this machine, the bar is passed through an automatic swaging machine. 4 PHILIPS TECHNICAL REVIEW VOLUME 25

Volume XXXIV. June, 1912. No.6

THE PHYSICAL REVIEW.

CONVECTION AND CONDUCTION OF HEAT IN GASES.

By IRV!NG LANGI\.1UIR.

PART I. HISTORICAL. HE loss of heat by convection from a heated body has apparently T always been looked upon as a phenomenon essentially so com- plicated that a true knowledge qf its laws seemed nearly impossible. A. Oberbeck' gives the general differentlal equations for this problem but finds it impossible to solve them for actual cases. L. Lorenz? for the ease of vertically placed plane surfaces is able to obtain some approxirnate

Fig. 3. Irving Langmuir (left), in conversation with Sir . .T. J. Thomso n, the discoverer of the electron. (The photo, taken in 1923, is by courtesy of General Electric Research Lnboratories Schenectady.) Right, the opening lines of the first of Langmuir's publications that lcd to the development of the gas-filled lamp.

so ductile that at high temperature it could be drawn search which fundamentally widened its potentiali- into wire of any required thickness, down to the very ties, may still be regarded as a classical example of finest. applied physics. The fact that this ycar marks the The Philips factories, too, very soon adopted this 50th anniversary of the gas-filled lamp has prompted process, and their first drawn-tungsten-filament us to review this interesting work once again. We lamp was made on 5th Deccmber 1911. Owing to shall first consider the invention itself, and then this great effort the new lamp was brought out by trace the subsequent development. Philips almost simultaneously with those of compe- Langmuic's invention ti tors. In July 1912 the production of squirted-fda- ment lamps was stopped altogether, and from then A critical study of experiments carried out by on only lamps with drawn filaments werc put on the Nernst conccrning the formation of nitric oxide on market. an incandescent wirc in air 2), induced Langmuir to The new technique using drawn tungstcn wire investigate the loss of heat by convection from a had hardly been introduced in the factory when a wire heatcd to incandescence in a gas. new discovery was announced ~ the incandescent The attraction of burning the filament in an inert lamp with a gas-filled bulb. That was at the begin- gas (i.e. one which would not react with the white- ning of 1913, now half a century ago. As the new hot tungsten) instead of in a vacuum as done pre- lamp consumed about half a watt per candle, it viously, was that the surrounding gas considerably soon became fairly generally known as the "half- slows down the evaporation of the tungsten which watt lamp". The credit for the invention of this is responsible for the blackening of the bulb. This lamp was due to the distinguished physicist hving made it possible, while maintaining the same useful Langmuir, who was working in the laboratories life, to heat the filament to a vcry much higher of the General Electric Company in Schenectady temperature, the higher the filament temperature (D. S. A.); seefig. 3. the better being the conversion of the electrical At first the invention related only to large lamps, energy into light. With a gas filling, however, heat of 600 to 3000 candles. The gas filling was not yet is lost by conduction via the gas. If nothing is done suitable for lamps of lower power, which claimed the about this, it will completely offset the gain of bet- lion's share of the production of the lamp factories. ter energy conversion. It was precisely the object of Intensive research throughout the world, however, Langmuir's investigation to reduce the heat losses enabled the major categories of these lamps to ben- 2) W. Nernst, Chemisches Gleichgewicht und Temperatur- efit from the same principle within a few years. gefülle, Festschrift L. Bol tzmann, published by Barth, The steps that led to this invention, and the re- Leipzig 1904, pp_ 904-915. 1963/64, No. 1 HALF-WATT LAMP 5 caused by the gas filling. His results appeared in a muir showed that B is proportional to the viscosity number of now famous publications 3) (see fig. 3). of the surrounding gas, inversely proportional to the We shall here very briefly summarize the results of density of the gas, inversely proportional to the his experiments and theoretical work 4). 0.75 power of the gas pressure, and finally roughly The viscosity of a gas increases with increasing proportional to the absolute temperature of the gas. temperature. In the immediate neighbourhood of At values of B and of the filament diameter a likely an incandescent body the viscosity is so high even that a gas no longer flows. For quantitative pur- '10 poses it is convenient to assume that every incan- s descent body in a gas atmosphere is surrounded by i--' a stationary layer of gas, whose outer surface has t ~ V roughly the temperature Tl of the ambient atmos- V" 3 V phere, and inner surface the temperature T of the 2 1-1""" incandescent body. 2 I---- '- - On the basis of this hypothesis, in which the incandescent filament loses its heat, apart from radi- 1 ation, solely by conduction through the stationary 0,01 0,02 0,04 0.06 0,1 0,2 0.4 0.6 0,8 1 layer of gas, Langmuir found that the heat loss W -a/B per unit length of filament (in Wfmin) can be defined Fig. 4. Graphic representation of equation (4). In practice al B varies only within the limits indicated along the abscissa. by the formula:

2:n; to be encountered in practice, af B is found to lie W = b (ep2 - epI) , (1) between 0.02 and 0.2. (We shall presently see why the In - a term "filament" is now used and not wire.) In this range of values the curve in fig. 4 can be represented where with reasonable accuracy by the equation: Ti epi = 4.19 J k dT ...... (2) 2: = s = C(i)O.3, . .. (5) ° In - Here a is the diameter of the filament, b the diameter a of the cylindrical, stationary gas layer and k the C being a constant. The usefulness of this formula coefficient of thermal conductivity of the gas (in was later confirmed by numerous experiments. calfm degree s). For calculating the factor For a filament of length 1 and diameter a (both in 2:n; mm) the total heat loss Wg (in watts) is now: b' (3) ln- a Wg = Cl (i)O.3(ep2 - epI) • (6) occurringin (1) and usually denotedby s, which con- It can be seen from this that the heat losses to the tains the unknown diameter b of the stationary gas gas are primarily determined by the length of the layer, Langmuir gave the formula: filament, while its diameter is of subordinate in- -2", fluence. For an ambient temperature Tl of 300 "K ass - - =-e (4,) and a filament temperature T between 2500 and B :n; 2 3300 "K, we can express ep2 - epI as a function of T2 where B is the thickness of the stationary layer of for the ga~es nitrogen, argon and krypton by: gas produced on an incandescent flat plate in the same gas. Fig. 4 shows a plot of s versus afB. ep2 - CPI = a (2;;of, (7) 'I'he: value of B was of fundamental importance for the later practical application of the results. Lang- in which the constants a and {J have the values stated in Table J. Fig. 5 gives a graphic representa- 3) I. Langmuir, Convection and conduction of heat in gases, Phys. Rev. 34, 401-422,1912; also Proc. Amer. Inst. Electr. tion of equation (7). Engrs. 31, 1011-1022, 1912. Idem, Convection and radiation From the foregoing it may be concluded that it of heat, Trans. Amer. Electrochem. Soc. 23, 299-332, 1913. 4) Many years later, experiments were done to determine the would be advantageous to have a short, thick fila- influence of deviations from certain simplifying assumptions ment, and from equations (6) and (7) one would introduced by Làngmuir; see 1. Brody and F. Körösy, J. appl. Phys. 10, 584, 1939: Further: W. Elenbaas, Physica then be able to calculate the heat losses and from this 4, 761, 1937 and 6, 380, 1939. the luminous efficiency. But if one wishes to make an

.. 6 PHILIPS TECHNICAL REVIEW VOLUME 25

Table i. transfer to the gas is concerned, like a short cylindri-

a atomic cal incandescent body with a diameter equal to the Gas (J in W/min weight outside diameter of the coil. A simple calculation, using formula (6), shows that this makes it easily Nitrogen 0.170 *) 1.65 14 possible to reduce the heat loss to the gas to 15%, Argon 0.122 1.65 40 which largely overcomes the drawbacks of the gas Krypton 0.073 1.65 83 filling - at least for lamps of high wattage. For smaller lamps, as will appear in the following, the *) This was calculated without taking into account the dis- balance was at first still unfavourable. sociation of the nitrogen molecules N2• Experiments show that the value, accounting for dissociation, is about 0.22. 100 incandescent lamp for a given voltage and power, bla the length and diameter of the filament are already so~ fixed. This brings us to the crux of the whole prob- <, i30 lem. After filling in (3), we can write equation (4) <, in the form 5): 20

b b (-1 0 -ln- = 2 !!..) . (8) "" a a B -, <,r-, Since al B, as mentioned, lies between 0.02 and 0.2, 5 we see that bla, the ratio of the diameter of the 3 " <, stationary gas cylinder to the diameter of the fila- 2 ...... ment, is for all practical purposes much larger than ......

1 (see fig. 6). If we wind a long thin wire into a 1 0.01 0,02 0,04 0,06 0,1 0,2 0,4 ~ 0,6 0,8 1 2 4 short helix, so that the pitch of the successive terms _a/8 is about 1.5 times the wire diameter, then the sta- Fig. 6. Ratio of the diameter b of the stationary gas layer to the tionary gas layers of the separate windings com- diameter a of the (cylindrical) incandescent filament, as a function of «[B, calculated using Langmuir's equation (8). pletely overlap each other. When coiled in this way the long thin wire thus behaves, as far as heat Realization of the invention To put Langmuir's invention into practice it was 0,20 I...... -L_ necessary to coil tungsten wire into a close helix. ~-I'} i"3----' ::l_ ,__ This called for ductile wire, so that Coolidge's pro- 10,16 cess for drawing tungsten wire had arrived just in ...... 0,14 V time...... + Ar_ ...... Production brought other problems, however, the 0,12 /' V V first among which was the blackening of the bulb - »> ...... V the very effect the gas fillingwas meant to overcome. 0,10 - r- ~ ",..... Years of experience had already been gained in ,..-/ combating the blackening of incandescent lamps. 0,08 »> ...... - Kr",..... »> In the manufacture of carbon-filament lamps it had V been discovered that the bulbs, and the glass ...... --- vacuum system used to evacuate the lamps, should 0,06 ...... --- be thoroughly dried. This was done by heating the .....- »> bulbs under the pump hoods to several hundred ~I-"" degrees eentigrade and removing the liberated wa- 004 ter vapour with phosphorus pentoxide. The residual • 2000 2500 3000 water vapour and the remaining air in the bulb Fig. 5. Plot of the quantity rp2-rpl in Langmuir's heat-loss were removed after seal-off: for this purpose a small equation (eq. 1) versus the filament temperature T2 in "K, re- presenting the theoretical values for different gases (in the case quantity of red phosphorus was placed in the ex- of N2 no acccount is taken of dissociation). haust stem and heated to evaporation while the 6) Langmuir really gives the relation: carbon filament was burning. a When the change-over was made from the car- bIn b = 2B, bon-filament lamp to the tungsten-filament lamp, which is not, however, as clear as (8). bulb blackening sometimes occurred to a very serious 1963/64, No. 1 HALF -WATT LAMP 7 extent. It was first thought that the black deposit With the gas-filled lamp the situation as regards consisted of carbon originating from the organic blackening is in itself much more favourable than binder used when making the squirted tungsten with the vacuum lamp: the hot gas near the fila- filaments; later it was found that the deposit was ment - outside the stationary layer - rises and not carbon but tungsten. Since the blackening varied carries with it the evaporated tungsten, which is from lamp to lamp, it was probable that it was not therefore mainly deposited on the bulb wall directly only attributable to the normal evaporation of above the filament. In lamps mounted in a hanging tungsten but also to a substance present in some position the deposit thus forms in the neck, where

Fig. 7. When evacuating the lamp bulbs on the pumps the last traces of water vapour are removed by a liquid-air trap in the pump line (see the Dewar flask roughly in the middle of the photograph).

lamps and not in others. It was assumed that water it is least troublesome. On the other hand, all glass vapour was again the culprit, apparently somehow parts ofthe lamp, particularly on the inside are heat- being able to accelerate the transport of tungsten ed by the hot gas to much higher temperatures than from the filament to the wall. When the temperature in the vacuum lamp. This apparently was the reason under the pump hoods was appreciably increased, why in the early attempts at production residual so as to release the vapour still present on the bulb water vapour was still being released during life, re- wall, and when this vapour was then exhausted, sulting once again in excessive blackening. It was the irregular blackening was in fact no longer found. now no longer sufficient simply to raise thetempera- 8 PHILlPS TECHNICAL REVIEW VOLUME 25 ture under the pump hoods during evacuation. A On 19th November 1913 the Philips factories better means of drying had to be found. Only when announced that they were now able to supply gas- liquid air was adopted as the drying agent was it filled lamps of 600 to 3000 candle-power (fig. 8).

NAAM LOOZE VENNOOTSCHAP PHI LI PS' M ETAAL-GLOEI LAM PEN FABRIEK.

TELEGRAM-ADRES : "META EINDHOVEN"

ABC CODE" EH hUlTGAVE. L)E.£R'8 COD£. MAST ER CO D E.

TEL INTERC. No. 92.

In uw antwoord _ .. rwfjön " .. r:

Ij ZEN

.rer

PHILIP~ ._BALi!'WATT" LAIlPEN.

60- 70 Volt 300 Watt N.K. PI. 8.76 p. 60~ passend in Armaturen 50- 70 0 0 ·...l at 600 1000 N.K. 10.60 PI P2 116-136

60- 70 1000 Watt 2000 N.K. PI. 16.16 p. stuk passend in 115-135 VO:lt ~ Armaturen 1500 3000 N.K. 21.-- 0 200-260 ~ PHI & PH2

A R M A.T URE N.

PI zonder reflector bestemd voor 600-1000 N.K. FI. 11.60 p. stuk. P2 met 600-1000 12.60 PHI zonder 2000-3000 14.50 PH2 met 2000-3000 16.60 • De armaturen z~n zwart ge~mailleerd met helder en of melkglasballon. Levering van lampen en armaturen FRANCO·.-RUIS incl. verpakking.

o N DER D EEL E N.

heldere of INo:LlvoorarmaturenPl &P2 230llj/mdoor-PL1.30p.st. melkglasballon INotL2 _ • PHI & PH2 300 ,snede 1.75 •

)l(o:Sl PI &pg 460 • 1.10" • ;110:52 PHI &PH2 550 • 2. -- • Onderdeelen af fabriek exclusief verpakking. Transportbreuk wordt uitsluitend vergoed b~ fran'qo terug- lending der lampen onmiddell~k na ontvangst.

Fig. 8. Price quotation of N.V. Philips' Metaal-gloeilampenfabriek, dated 19th November 1913, with the first offer of gas-filled lamps. As can be seen, the lamps were rather expensive: approx. Fl. 10 ("'" £ 1 today) per lamp. The lamps for 2000 and 3000 candles could be made for voltages up to 250 V, lamps for 600 candles only for the lower voltages of 50 to 70 V. possible to reap the benefits of the gas filling, which This was hardly a month after Langmuir and J. A. are due to the reduction of the rate of evaporation Orange had presented a paper in New York to the of the tungsten (fig. 7). Americal Institute of Electrical Engineers, in whieh they reported on the realization of the gas-filled Blackening is still one of the fundamental problems in the lamp 7). construction of i ncandescent lamps - particularly in connection with bulb dimensions, as we shall presently see. This is demon- The lamps of the candle-powers mentioned were strated by the reeent development ofiodineincandescentlamps, made for the lighting of streets and large enclosed where the problem is t.ackled again 6). 7) 1. Langmuir and J. A. Orange, Tungsten lamps of high 6) See e.g. J. W. van Tijen, Philips tech. Rev. 23, 237, efficiency, Proc. Amer. Inst. Electr. Engrs. 32, 1893-1926, 1961/62. 1913. 1963/64, No. 1 HALF- WATT LAMP 9 spaces; they were competing here with the carbon- For a given power at a lower voltage a thicker and are lamp, which was in use for these purposes. Are shorter filament is needed. For this reason Philips lamps burnt very irregularly and called for a great introduced in 1914 lamps of 100 candles for a volt- deal of maintenance, so that it is not surprising that age of 14 V. Here the filament could again be coiled from then on they were gradually superseded by and a gas filling used with advantage. A gas-filled the gas-filled lamp. lamp of low power was thus obtained, but a trans- The fact that the gas filling was at first used only former was necessary in order to use the lamp with in high-power lamps may be understood as follows. the normal mains of 220 or 110 V. In 1914 part of For a higher power at a given voltage, a thicker and the Kerkstraat in Amsterdam was lit by 21lamps of

Fig. 9. Fitting containing a lOO-candle gas-filled lamp, part of the lighting installation in an area of the Kerkstraat in Amsterdam in 1914. Each lamp required its own step-down transformer for 220/14. V. (Photograph by courtesy of the editor of "De Koppeling"; see also that journal, vol. 8, 149, 1953.)

longer filament is needed. Such a filament can be this kind, each lamp being provided with a 220/14 V wound on an appreciably thicker mandrel than the step-down transformer (see Jig. 9). Although the thin wire for a smaller lamp without the coil be- Municipal Electricity Corporation was well satisfied coming too limp. Consequently the length of the with this installation (see the extract from a report coil in both cases can be roughly the same, only the reproduced in Jig. la), it was not a satisfactory so- thickness of the filament being greater. Since, how- lution for private users. ever, the heat transfer to the gas increases, as ap- Again in 1914, Philips announced a 200-candle pears from formula (6), by only the 0.3 power of the lamp for 220 V and a lOO-candle lamp for 110 V, filament thickness, this loss is of much less impor- and as early as Lst November of the same year the tance in large lamps than in small ones. price of these lamps was drastically lowered (Jig. 11). 10 PI-IILIPS TECHNICAL REVIEW VOLUME 25

This rapid improvement was partly due to the difference between monatomic argon gas and dia- introduction of argon for the gas filling instead of tomic nitrogen gas does not seem very considerable the nitrogen originally employed. It was evident (respective molecular weights 40 and 28), but an that gases of greater molecular weight would be undesirable effect of nitrogen is that it dissociates at more suitable because of their lower coefficient of high temperature, so that in fact nitrogen compares ~~.L.3 ~:.' NAAMLOOZE VENNOOTSCHAP LI PHILIPS' GLOEILAMPENFABRIEKEN r ./'

..PHILlPS EINDHOVEN"

"- B. C. 0001 "I' eN S- UITe"" ...... LI •• l.n COOL "1:~~EINDHOVEN. Oe\obfJr 19111. m.i''''~''.''.-.. ~G~~ j) r lol ••

lJjJ~~~en.D aan het ~g. betreffende de

Electriciteita Werhen te Amaterdam:

,De gemiddelde brandduur voor de b~ de openbare ,verliohting in gebruik z~nde Halfwattlampen ,bedroeg ongeveer voor: 1914 i o i e

Lampen 100 N.K. /14 Volt 1600 uren 1700 uren r 200 N.K. /220 1100 1'100N.K. /220 7Bl 1200 fJA. ¥ ~.,/!1(IrJ--rJ 1000 N.K. /220 1340 1366 ./~I"_4Vl'v:; 1600 N.K. /220 86B 1247 1~£

geschiedt met circa 400 Philips Halfwattlampen van versohil-

lende liohtsterkten. De hlerbovengenoemde o~fors spreken voor

3ioh self.

Het ie ona aangonaam, U dit ongevraagd rapport ter

kennismakin& t. hunnen toezenden. Bij best.lli~. van Halfwatt-

lampen lette men 01' DEN STEIlI'BL ,PHILIPS". Hoogachtend. N.V. PHILIPS' GloGilampenfabrieken. H 211

Fig. 10. Copy of an extract from an annual report for 1916 relating to the Amsterdam electricity works. The extract shows that in 1915 Philips were already supplying 200-candle (N.K.) gas-filled lamps for 220 V, which lasted for 1100 hours. The report also states that the complete (electric) lighting of Amsterdam was done with about 400 Philips half-watt lamps. The handwritten part adds that about 7000 gas lamps were also used.

thermal conductivity k (see equation 2). Subsequent even more unfavourably with argon than was pre- investigations, including work done in the Philips dicted in the theory (see the values of a in Table I). Research Laboratories, which had meanwhile been The use of argon therefore considerably improved established 8), showed that a gas of greater molecular the heat balance. weight has a further advantage in that it reduces As air contains a relatively large percentage of even more the rate of evaporation of tungsten. The argon (about 1%), it was possible to use this inert gas on a large scale. The argon was supplied to Phi- 8) E. Oosterhuis, Chem. Weekbl. 14, 595,1917. See also W. Geiss, Philips tech. Rev. 6, 334, 1941. lips by a German firm, "Gesellschaft für Lindes 1963/64" No. 1 HALF -WATT LAMP 11

EINDHOVEN, I 'November 15!1i.

Belangrijke Prijsverlaging.

Philips' ,,112 Watt" Lampen van 100 Kaarsen

voor Winkel-, f;talage·, Restaurant· en Huisverlichting.

Door de zeer groote vraag naar PHILIPS' 'j, WATT ,LAMPEtJ in kleine k~ (.lj.a wij tot massa-fabncaqe kunnen overgaan en wenschert wij de daaraan verbonden vooedeelea III .t)e,n vorm eener belangrijke' prijsveriagiDg den verbruikers ten goede te doen komen, De prijs der 100 N.K. lampen. is gebracht van f i,20 op f. 2,70 per stuit. Ortderstaande stroomberekening bewijst Uw voordeel bij het gebruik van deze lampen. 125 VOLTS 100 KAARSEN 110 VOLTS f 2,70 per stuk.

De stroomkosten van 2 metaaldraadlampen 110 of De stroom kosten van EENE .'J, Watt" !mI' 125 Volts, 50 Kaarsen bedragen bij een tarief van 110 of 125 Volts, 100 Kaarsen bedragen bij eea 20 ct, per K W.U, en een gemidd'elden brandtijd tarief van 20 ct. per K.W,U. en een gemiddeldtn van 1(){X) uren per ·seizoen en per lamp: brandtijd van Jooo uren per seizoen en per lamp: 1000 X 2X 50 XII, ,= 1\0 K.W,U. 1000 X 100 X 0,6 . . . = 60 K.W •.U. ad FI. 0,20 , =' FI. 22,- ad FI. 0.20 . '. .. = FI. 12._ Lampenverwtsselinq na gemidd~d I Lampenverwisselinq na gemiddeld 1000 uren 2 X Fl 0,55 , , -" 1 10 800 uren:' 'I. X' Ft. 2.70 _=_:",,,_3_:J1~_ Totale verlichtingskosteD.. FI. 23..10 Totale verlichtingskfsten 'FI. {5.37

BENE TOTALE BESPAI.ij'HG DUS VAN. PLo 731.

Spaart dus stroom en geld en vervaagt Uwe' metaaldraadlampen. dobr:

PHILIPS' "'j, WATT" LAMPEN VaD 100 KAARseN. Leverlnjj geschie9t uitsluitend door bemiddeling ~an H,H. lziatailllteurs ea Wllllerv~.

Hoogacbtead. N.V PHlLIPS' Melaal-Glodl~

Fig. 11. Announcement of lOO-candle gas-filled lamps at substantially reduced prices JIl November 1914: price of a lOO-candle lamp dropped to Fl. 2.70.

Eismaschinen" of Höllriegelskreuth (near Munich), the gas being a byproduct in the preparation of oxygen and nitrogen by the fractional distillation of liquid air. The outbreak of the first world war, however, quickly put an end to these supplies. Fortu- nately, a member of Philips' staff at that time had acquired considerable experience in the liquefying of inert gases, and within a remarkably short time the Philips factories were consequently able to build and operate their own fractional distillation equipment, thus ensuring sufficient argon for their .requiremen ts. An important and at first sight perhaps unexpect- ed consequence of the gas filling was that lamps could be made much smaller for a given power Fig. 12. The compact form of the filament and the favourable (fig.12). This is bound up with the fact, already men- distribution of thc gradually forming black tungsten deposits made it possible to produce gas-filled lamps (left) with a much tioned, that the evaporating tungsten is now carried smaller bulb than vacuum lamps of the same power. 12 PHILlPS TECHNICAL REVIEW VOLUME 25 upwards by the rising gas, and settles almost entirely This problem led to intensive research all over the in the upper parts of the bulb. In the vacuum lamp world to find a filament that would not sag - espe- the tungsten is deposited all over the bulb wall, and cially after it was found that tungsten containing therefore this wall must be given a large surface certain impurities showed far less tendency to sag area to ensure that the tungsten deposit remains than completely pure tungsten. sufficiently transparent. In this case, in fact, the When pure tungsten is used as the starting ma- usefullife is governed by the increasing blackening; terial in the manufacturc of the filament wire it is on the other hand the useful life of the gas-filled lamp is limited by the occurence of thin or weak spots in the filament after a certain amount of tungsten has evaporated, causing the tungsten wire to break. Consequently, the choice of bulb diameter a of a gas-filled lamp does not dcpend on blackening and usefullife, but can be made as small as the temperature of the bulb wall permits (the bulb of course becoming hotter as its diameter decreases). When much later, with a view to the further improvement of luminous efficiency, the even heavier b inert gas krypton was considered as the filling gas instead of argon, the high price of krypton made the bulb volume itself an important consideration. On Fig. 13. a) Structurc of a wire of pure tungsten after recrystalli- this subject reference may be made to the article by zat.io n. Geiss quoted above 8). Similarly, the change brought b) "Offsetting" in a tungsLen fi la mnnt., occurring when a larn p having a filament with the structure shown in (a) had burnt for about in the radiation properties of thc filament as a some time. result of coiling will not be dealt with here. This too was the subject of cxtensive and much morc rccent scen after recrystallization, which occurs at a high investigations, including joint research carried out temperaturc, to givc the wire a structure as shown by the Osram and Philips factories 9). in fig. 13a. After thc lamp has burnt for some time, vibrations and the force of gravity cause sliding Further developments, notably of the filament along the boundaries of the tungsten crystals, pro- Although the drawn tungsten filament possessed ducing the effcct called "offset.ting " (fig. 13b). This very good mechanical propertics and could readily effect promotes the sagging of the coil. Since it also be coiled, an unforeseen difficulty in the early days promotes local irregularitics in the filament tempc- of the gas-fillcd lamps was the tcndency of the tung- rat.ure, it has a disastrous influence on the life of thc sten filament to sag. At thc very high tcmperature lamp. to which thc coil is heated in operation the metal In the early development of the gas-filled lamp became so soft that the windings of thc coil grad- the Philips factories prepared tungsten by the ually opened out anel sagged under their own weight, "Battersea process". In this process tungstic acid sometimes even resulting in an almost straight wire. was heated in closed, refractory crucibles at high

Owing to the lengthening of the filaments which temperature, about 1200 °C, removing the H20 and accompanied this sagging, thc heat losses increased producing a very compact W03, which was then and the light output therefore dropped sharply. If reduced to tungsten in the normal way. The WOB

the advantages of the coiled filament are to be absorbed impurities from the crucibles, mainly K20,

maintained during the whole life of the lamp, the Si02 and A1203• The presence of these impurities spirals must keep their original shape as far as evidently produced a different recrystallization possible. This is indeed still a problem calling for texture from that in pure tungsten wire. Fig. 14 constant attention, particularly in lamps of smaller illustrates this texture. wattage 10). The quantity of impurities taken up varied considerably, however. By mixing portions of tungsten from different crucibles it was possible to arrive at a certain favourable quantity. It was fortunate for 9) G. Holst, E. Lax, E. Oosterhuis and M. Pirani, Leucht- dichte und Gesamtstrahlungsdichte van Wolfrarnwendeln, our work that this insight into the behaviour of Z. techno Physik 9, 186-194, 1928. tungsten was already available in our laboratories at '0) See e.g. E. W. van Heuven, Shock testing of incandescent lamps, Philips tech. Rev. 24, 199-205, 1962/63 (No. 7). the very time we were developing the gas-filled 1963/64, No. I HALF-WATT LAMP 13 lamp. The Battersea wire could be used with good before the reduction process - but none of them are results in the gas-filled lamp, and the sagging of the entirely satisfactory 12). coils was thus kept within reasonable bounds. The argon gas offered, as described, the advan- Even better results were obtained by applying an tage ofremoving less heat from the filament than ni- American process 11) in which, before the reduction, trogen. A disadvantage, however, was that its a controlled quantity of Na-K silicate was added to breakdown potential was considerably lower than the tungstic acid. In this way tungsten wire was that of nitrogen. Consequently, if the lead-in wires came too close to each other, arcing occurred between them, prematurely ending the life of the lamp. For- tunately it was found that this effect could be sup- pressed if a certain percentage of nitrogen was added to the argon (about 10%), provided the argon pressure was not too low. This breakdown problem was primarily encountered in the Euro- Fig. 14. Recrystallization texture of a tungsten wire containing pean countries, where the mains voltage is generally impurities resulting from the Battersea process. higher than e.g. in the U.S.A. As time went on, more and more types of lamps with argon filling appeared, and they were produced in very large quantities. Tungsten-Jamp manufacture in the 'twenties was therefore characterized by increasing mechanization of the production process. In the 'thirties a further important improvement Fig. IS. Recrystallization texture of "doped tungsten", i.e. tungsten containing controlled amounts of additives. The was made to the gas-filled lamp. In Europe the stand- elongated, wedge-shaped crystals give the wire considerable ard tungsten lamps for voltages above 200 V have strength and prevent sagging when thc wire is coiled. a long filament wire, and even when coiled the fila- obtained which, after rccrystallization, consisted of ment is still relatively long. With the idea of re- long overlapping crystals; see fig. 15. This process ducing the cooling area, it was decided to coil thc was soon adopted in the Philips factories. With the filament doubly(fig.16), thus producing the "coiled- new wire all sagging difficulties were overcome, coil" lamp. This resulted in a quite appreciable

Fig. 16. Coiled-coil filament for a 100W lamp. The inset shows a piece of the coiled-coil at higher magnification.

making it possible later also to meet the reqUIre- improvement of luminous efficiency particularly at ments of the coiled-coillamp (see below). lower wattages 13). The coiled-coillamp for low watt- Many theories have been put forward to explain the action of the additives ("dope") - N.B. added 12) See e.g. J. L. Meijering and G. D. Rieck, The function of additives in tungsten for filaments, Philips tech. Rev. 19, 109-117,1957/58. H. L. Spier, Influence of chemical additionson thereduction 11) U.S.A. Patent 1410499, filed Feb. 1917, granted March of tungsten oxides, thesis Technische Hogeschool Eindho- 1922 in the name of A. Pacz. ven, 1961. 14 PHILlPS TECHNICAL REVIEW VOLUME 25

ages is in fact a specific European contribution to the development of the incandescent lamp. The coiled-coil lamp makes even more stringent demands on the gas filling and on the non-sagging properties of the filament than the single-coillamp. A coiled-coil can only be made with the best non- sagging wire. It is produced by first winding the tungsten wire on a molybdenum wire mandrel of the appropriate thickness, and then winding the coil thus obtained, together with its mandrel, around a second thick molybdenum mandrel. After heating the filament to incandescence for some time to "set" the wire, the two mandrels are removed by dis- solving them in a suitable acid. A good solution was found in the Philips factories for the technological problems which this involved. Finally, to illustrate the achievements to date, fig. 17 shows the luminous efficiencies (in lumens per watt) of three types of lamps manufactured today: vacuum lamps, single-coil gas-filled lamps, and coiled-coil gas-filled lamps. All three curves relate to lamps for 225 V having a usefullife of 1000 hours. It can be seen that the gas filling, in conjunction with a coiled filament, is now used with advantage for lamps of 4,0Wand higher. Also clearly to be seen is the gain obtained by coiled-coil filaments 13), particularly at thelowerwattages. The name "half-watt" for the two latter categories of lamps is not properly relevant for lamps given on this graph. Broadly speaking a luminous flux of 10 lumens is equivalent to a luminous intensity of one candle. Thus it is only at luminous efficiencies of about 20 lm/W, encountered when extending the graph to 1000 W, that one has half-watt lamps. 17/_r!>_w 16 V The combination of a gas filling with a coiled fila- 15 ./ --- ment proved to be desirable also for tungsten lamps 14 G-?-, , - / - - other than those used for general lighting, and in IJ ,- - V'" , fact nearly all types of vacuum lamps were replaced 12 , / ... by gas-filled types. Indeed many kinds of lamp ... ' 'G1 11 ,- only became possible because coiling allowed the // ,, 10 , construction of a sufficiently compact filament. One ,, V_""""- of the most striking examples is the tungsten lamp 9 ~ , V for film or slide projection. The very compact fila- 8 ..-/ ment provided the necessary high average luminance, V while the possibility of making the bulb very small 7 was essential to the effective design of the opti- cal system and for limiting the size of the whole 10 15 25 40 60 75 100 150 200W projector. The same factors were decisive in the de-

Fig. 17. Luminous efficiency of present-day vacuum lamps (V), velopment of special lamps for car headlights and of gas-filled single-coil lamps (G,) and gas-filled coiled-coil of many other similar types. The invention of the lamps (G ), as a function of wattage. 2 gas-filled lamp has therefore contributed in no small measure to the extraordinary diversity of incandescent lamps now available; at the present time, for 13) W. Geiss, On the development of coiled-coil lamps, Philips tech. Rev. 1, 97-101, 1936. example, Philips produce some tens of thousands of 1963/64., No. 1 HALF- WATT LAMP 15

Fig. 18. A small selection from the tens of thousands of types of incandescent lamps nowa- days produced by Philips. Most types come into the many categories of "special lamps" (projection lamps, car bulbs, window-display lamps, airfield, sports-field and lighthouse lamps, studio lamps, infrared-heating lamps, signal lamps, bicycle bulbs, etc. etc.); but a large number of types also come into the category of standard lamps for domestic and street lighting, made with numerous variations of wattage, voltage, kind of bulb, etc.

types. To conclude this review, our last figure is given of the theory underlying Langmuir's invention. The problems involved in the manufacture of gas-filled lamps are (jig_ 18) shows a small selection from this enormous discussed (named half-watt because the power per candle was variety. reduced from about 1 W to about t W - at least for lamps of more than 2000 candles), in particular the blackening of the Summary. The first gas-filled incandescent lamps appeared at bulb and the sagging of the coiled tungsten filament. A con- the beginning of 1913, now half a century ago. The invention cise review is given of subsequent developments, which led to of the gas-filled lamp, which followed from the work of Irving the use of inert gases for the gas filling, and to the invention of Langmuir, is recalled in this article. After a brief history of the the coiled-coil filament, which allows the use of a gas filling even development of the incandescent electric lamp, a short account in lamps of relatively low wattage.