I N D U S T R I A L A N D E N G I N E E R I N G CHEMISTRY
1512
adeauate coverage of estensive sample breakdown. Under the. conditions emploJ-ed, sulfuric acid appears to be only about one sisth as active in hydrolyzing cellulose as is hydrochloric acid of equivalent concencration. The data suggest that cellulose molecule sections which intcrlink the crystallites in the chain direction are attacked first. This part of the structure, highly hygroscopic and probably quite disorganized, represents relatively m a l l percentages uf the intact materials. Hydrolytic attack thereaftcr appcars t o bo limitcd to lateral crystallite surface.;. Crystallite length estimaks derived from from 280 glucose units for cotton to 110 for rayon. Mercerization apparently causes a lite length. I
Vol. 39, No. 11
descrilxd in this series of paners. and the kind assistance of L. Bass, G. D. Beal, K. A. Hamor, and W. H. Child of ilon Iiistitutc in the preparation of manuscripts. I
I
ACKIVOWLEDGMENT
made by the Satiorial Cotton Council for the invcstigiitions
I
~
~
:
~ 3, ~ 19-16 ~ \
-
~
~
Mean Molecular Weights of Asphalts and Their Constituents GEORGE IT. ECICERT AND BRUCE WEETMAS The Texus Gonipany, Beacon, S. Y .
T h e molecular weight method based o n the Tiscosit) of dilute solutions was investigated for its applicability to asphalts and their constituents. The two equationi proposed by Staudinger for calculating molecular w eights from Iiscosities of solutions w ere considered. The constants in the equations were etaluated bj using cryoscopic molecular weights for t h e oil and resin fractions from asphalts. The molecular weights of t h e oils and resins ranged from 370 to 900. Viscosities of benzene solutions were determined a t 7 i b F. i n a capillary t j p e tiscometer. RIodifica-
tions of Staudinger's equations are not required for theoil arid resin fractions of asphalts 011 the basis of the present ~ o r k>lean . molecular weights of 700 to 1800 were obtained 1,) the ~ i s c o s i t )inethod for asphaltenes from asphalts of different source5. Both equations proposed b) Staudinper were used to calculate the molecular weights of asphaltenes, and neither appears preferable. Reliable independent data for the molecular weights of asphaltenes are not atailable to define conclusiYe1) t h e constants i n the original and modified forms of Qtaudinger's equations.
D
thc valucs obtaincd with benzene too high and ntt rihutcd incunresults t o ,separation of thc benzene solution into t \ w a t the freezing point. Hillnian and Barnc'tt (4)determined lar w i g h t s of asphalts by the cryosropic method, using both berizcne and n a p h t h a h e , antl obtained valiws for asphaltcncs from iioncraclred residua of 2300 and 1660 n i t h the two solw u t s , respoctiwly. They concluded that, the average molmxlar w i g h t o i ill-dcfined mistures, such H S asphaltenes, cannot he accepted as absolutc, sirice th(x solutions of these fraetioni proliabl?; dopitrt apprc'ciably f r o m the ideal solution Ian-. Grader ( 3 , t gives incorrect rcsults even d naplithalene as a solvent. molecular weights of asphaltencs by the Langmuir monomolecular filiii method and reportcd the following values:
URIKG studies in this laboratory on the properties of charac-
teristic components isolated from asphalts-namely, asphaltenes, resins, paraffin oils, naphthene oils, and waxes (fS),it has been of interest t o obtain the mean molecular neights of th(,se fract,ions. There is general agreement in the literature regarding the ordcr of magnitude of the mean molecular weights of t h r oil and resin type fractions. However, no such agreement exists for thr asphaltenes, probably because of their l o ~ vsolubility, eomplcx molecular structure, tendency to form aggregates, colloidal nat u w , and the fact t h a t their solutions may depart appreciably from the ideal solution Ian-s. I n view of the confusion esisting i n the literature, some attention has been given t o possible mcthods for clarification in regard t o the mean molecular w i g h t s of asphaltenes. Sakhanov and Vassiliev (14, reported molecular weights of 5000 t o 6000 for asphaltenes determined by the cryoscopic method using benzene. Strietcr (16) used bcnzcne for the cryoscopic molecular weight determination of asphalts and reported the following values: Trinidad bitumen 1131.8, Bermudez bitumen 620.3, gilsonite bitumen 4251.5. Katz ( 7 ) employed the cryoscopic met,hod,with camphor and brnzene, t o obtain the molrcular weights of asphaltenes. He obtained values of 4300 t o 5600 with benzene and 2219 t o 5160 with camphor solutions. He considered
1
Source of .\;phnltenes M e x i c a n , stearn-refined Alexican, blown Venezuelan. steam-refined Tenezuelan, b l o w n
M e a n l l o l e c u l a r Weight 80,000 80,000 110,000 140,000
Sivmison ( 1 : ) carried out diffusion experiments in which the h u n d a r y wts follon-td by light absorption, and claimed t h a t the rcijults suggest molccular w i g h t values for asphnltcnes at least as high as those obtained by Pfeiffer and Saal. Kirby ( 8 )used an isotonic or equal vapor pressure method ( I O ) m d obtained molecular weights of 1510 for gilsonite (select,) a r d 715 for a Nexican as--
November 1947
1513
INDUSTRIAL AND ENGINEERING CHEMISTRY
0.12 0.10
2 C
0.07
u 05 1 2 3 4 . 5 6 7 8 Concentration, Gram. >olute/100 1\11 Benzene
9
10
temperature belox which the polyniers break wddenly under increasing tensile load Kith no preliminary stretching. Above this brittle point, linear polymers stretch before breaking. Asphalts behave in this manner and may, therefore, comprise a series of compounds having properties resembling those of linear type molecules. Blthough it is higG1y improbable t h a t the fractions from asphalts are members of a homologous series, the conventional viscosity-molecular n-eight equations were used to evaluate This applicability of t h e viscositv of dilute polutions to asphalt components. TITOequations were proposed by Staudinger to relate viscosity of dilute solutions to the molecular Tyeight of the solutes:
41
Figure 1. Variation of q c with Change in Concent ration phalt. This implies t h a t the asphalterles have molecular weights in t h e range 2000 t o 3000, since gilsonite (select) contains 50 t o 60% asphaltenes. Molecular weights less than 2000 for asphaltenes were obtained b y Mack (11), employing viscosities of sohtions of asphaltenes.
;If
=
qsp/cKni
(1)
Ln log 77 'C
(2)
=
n here Jf = molecular m i g h t T a p = specific viscosity
= (qr
-
1)
(\ ibcosity of solution) ,'^
qp
=
relative viscositv
=
~
x,._. iacosity of solvent)
C = concentration K m , K,,, = constants
T h e published ~ o r k t,h e n , gives t\yo xidely divergent conclusions on t h e order of magnitude of the molecular weights of asphaltenes. High values of about 100,000 and Ion- values of about 3000 were reported. After the results of other investigators n-ere reviewxl, as well as early experiences in this laborator)., it was decided t h a t methods used for detc.rmining the moleculai, w i g h t s of high-molecular-\~cightpolymers might apply t o asphaltenes. T h e viscosity method has been discussed extensively in current literature in connection x i t h polymers. This method appeared to have certain advantages if it could be applied t o asphalt constituents. Thrrefore, ~ o r was k carriid out t o ascert'ain the applicability oi this method t o oils, m i i i s , and asphaltmcs iron1 asphalts. VISCOSITY-AIOLECULAR YEIGIIT EQUATIONS
Thc. limitations of the viscosity method for niulcwilai' weight determinations as sct forth b y Staudingc~r ( 1 5 ) are: ( a ) products must he members of a polymeric homologous series having a linear structure and a relntivcly high molecular weight; ( b ) tht. viscosity measurements must be carried out in a honiopolar solvent, a s it is only under this condition that the solute mol(w1es are surrounded by a monomolecular layer of solvent molccul(,s; and (c! viscosity measurements must be made in dilutib solutions t o avoid the gel state of the molecules in solutions. Itenis b and c are a matter of choice of solvent and concentration, and offer no difficulties with asphalt constituents. Item a is the limitation to which asphalt 'components may not comply, siiiw the nature of the components is still a matter of conjecture. Fuoss ( 2 ) stated that linrar polymers po ss a brittle point, a
1 2 3 4 5 6 7 8 9 Ciiiir entrntion, (>ranis Solute/100 1\11 Benzene
Figure 2.
4 086 1.083 0 0208 0.866 G 451 1,159 0 0216 0.993 5 000 1.134 0.0268 1.092 D (resin fraction) 7 152 1.235 0.0329 1.282 E (resin fraction) 3 0i8 1.096 0.0312 1.293 F (oil fraction) 3 466 1.107 0.0309 1.272 G (oil fraction) 3 760 1 118 0 0314 1.287 H (oil fraction) 10 809 1.386 0.0357 1.312 I (oil fraction) 2 71G 1.089 0 0328 1.362 J (oil fraction) 4 752 1,160 0 0337 1.335 ' K (oil fraction) 6 009 1 244 0 0406 1.580 L (resin fraction) 419 1.239 0 0411 1.718 11 (resin fraction) c> 123 1 225 0 0139 1.71Y S (resin fraction) 5 106 1 254 0.049i 1.925 a These fractions represent constituenrs which h:1r.e been isolated from a s p h a l t s ; t h e rnethc>d of sepal.ation will b e described i n a subsequent paper.
0
1
Variation of lop q l ' c with Change in Concentration
Huggin-: (5) stated t h a t the u w of Staudinger's equations t o obtain molecular weights of high polyinrrs is theoretically and experimentally unjustifiable, for most polymer-solvent systems at least. H e suggestcd the equation [q] =
KJlv
(3)
a s being more satisftictory for rctlatirig intrinsic viscohity of M solution of a polymtr t o t h o molecular n-eight of thcx polymer:
I:[
171 = intrinsic viscosity =
I (oil fraction) B (oil fraction) C (oil fraction)
1
e = 0 =
]p:
c = 0
Equation 3 was also proposed by X a r k (12) for the calculatiori of molecular w i g h t s of polymer5 from viscosities of dilute solutions. Flory ( 1 ) reported that the molrcular w i g h t s of polyisobutylenes calculated from Staudingcr's r.yrntions are too low. er, hc obtained a straight-linc relation by plotting the logarithm of q against t h e logarithm of the mdecular weightE (osmotic pressure) of polyisobutylmrs for the mdecular \\-eight range 6000 to 1,300,000. Thus, Huggins, LIark, and Flow found t h a t a modification of Staudinger's equations n-as gc-nerally suitable for polymi,rs of high molecular weight. Iiemp and Peters (8)shon-(d t h a t q..,,!c varici. n-ith molecular weight for some types of materials, and t h a t for these materials log qi'c is constant over a fairly n-ide range of moIecular w i g h t . I n attempting to apply the viwosity method t o asphalt constituents, the conventional equations as given here n-('re employed with thc Following results.
.
INDUSTRIAL AND ENGINEERING CHEMISTRY
1514
\'ISCOSITP L~I;TER~IINATION.The kinematic viscositie6 as described in A.S.T.11. Method D445--42T were determined 011 benzene solutions of the asphalt constituents. The viscosities were nieasured at 77 O * 0.05' F. in an Ostwald typr viscometer having a capillarj- of 0.6 nim. inside dianirter. The time of flow t'hrough the capillary ranged from 80 to 120 seconds foi, solutions having concentrations of 0.5 to' 3.0 grams of asphalt coiiarituent per 100 nil. bi~nzcnr. Thv viscositv of tht, heriwnt, n-a> 0.698 centistoke.
0 054 0 050
polatcd frciczing point for intiiiitv dilation ivas used for calculating molecular weight. Table I lists molecular w i g h t , q?,,"c, and log qr 'c for the constituents investigated. hsplialtt~nc~ fractions were not used for evaluating the coiistant bt>cauwt lit, cryoscopic method gives questionable values, as previously d i w w t d The constituents listed in Table I were sel(,ctcd twcause they cover a range of molecular weights from 370 t o 900 and includr variations in molecular types of matcri:ils. The 370 v;iluc~appmiches thc lowest probable molecular ncight obt a i t d ~ l sinw c t mphalts are distillation residues; !I00 is closc to til(' niaximum valuc that can Ire reliablv deterniincd 1)). the c.ryosc(Jpir iii('th(Jd f o r asphalt components. Tlica values i i i Table I for v,,,' P for the oils and rc,sins are plotted :qgii1is1 vryoswpie tiiolrvxlai, xvt'ights in Figure 3. The laotoi qFI, c I: a linwr function of this c.ryow)pic niol(wdar weights of thv oil> niid rcsin.;: I
U 046
Vol. 39, No. 11
.
0 042
[?Ic Figure 4 is it plot of [log q 7 '(;Ir for the oils and resiris against cryoscopic iiiol(-dar wi~igirit. A linear rtilatiori csspi,cw.od by t,he iollorving cqu:it ion \ v u ohtaincd:
0 032 0 028
0 024
(5)
0 020
400
600
800
900
hfolecular Weight
Figure 3. Relation of [q.,/c], of 4sphalt Constituents to l\Iolecular Weight
EFFECT OF COSCUSTRATIOS. For adaptioii of Equatioii 1, 2, or 3 t o the calculation of molecular weight, it is essc1riti:rl to kiioxv thc. effect of change in concentration on the factors q.,,, c and log qr /c. nd log qr!c for oils, resins, and asphaltrntxs arc in Figures 1 and 2 . Conc~rntrationc is tyiresscd as grams of solute per 100 ml. of benzeiw. The q.,,/c anti log qr!c values for a n oil fraction from a n asphalt tend t o change only slightly with increase in concentration. Therefore, finitc. valuc:: of qsp,!c and log 9 J c should be as suitablc as intrinric viscosity [ q ] for calculating molecular weights of tlic oils b y thcl viscositJmethod. The [q,p,!c]cand [log qr.,!c],.values for thc resins tend to increase slightly with increase in concentration. E tion of the curves t o zero concentration gives t h r intrinsic [ q ] . For concentrations less than 2 gram. per 100 nil., finitt. values of qsp,!cand log 9Jc are within 4?;. of [ a ] , Tht, s l o p c ~of the curves for the asphaltenes are greatc3r than thosc for tht, rtlsins. For asphaltene concentrations less than 1.0 gram ptsr 100 ml.. [log qr./clc is \Tithin 5y0of [log qr/c], = (I. Finite valurts of q q L , i c a t concentrations less than 1.0 gram pctr 100 nil. are within loci of [ 9al,,!~] e = 0 values. This would indiratc that [log q 1 'e] is i1rc.ferable t o [?s.p./c],; in other words, Equation 2 should gin. less error than Equation 1 for calculating the, molecular iveighth of asphaltenes from viscosity data. E V A L U A T I O S O F COSSTASTS. T o evalll:ttt, tllc COllstalltS for ii particular homologous series of polymers, molecular \vrights tictermined by an independent mc,thod on lower members of tht. series are substituted in the equations. For asphalt conatituents the structures are unknoxvn, and the possibility of selecting known members of a honiologous sc:~iesis cscludcd Thc altcrnativc remaining is to use t,helo\ver-moleculai-\vi,ight fractions of asphalt to evaluate the constants in the viscosit y -iiiolcndai~weight equations. The oil and resin fractions froin wphalts constitute thc lower-molecular-weight members of asplialt cvmponents. Thca molecular weights of these oil and rosin frart ions \vert obtained by the cryoscopic method with benzttne as solvent. The freezing points rrere determined at sewral concc~ntrations. and the estya-
h t l i Lquatioris 1 nlid 2 as givctn by Staudinycsr are applicahle to the oils and wsiiis investigated, I n regard t o Equation 3, it has alrt~adybeer1 explainid that for oils anti resins finite values of
or log q;:'c do not deviate appreciably from intrinsic viscosity nd log q, )zarc' linear functions of M, and there, Thus, for oils and resins fore riitis1ant Y in Equation 3 i froin asphalts, this conventional lwular weight equaI ions do not wquire niodificat ioii on t of tho prcwnt dat,a.
q",) 'C
0 021 0 019
0.01i
['?-I
c
0 (112 11
010
0 008
600 ,\Iulecaltlr Weight
400
800
900
Figure I.. Relatiori of [log qr'c], of isphalt Constituents to \Iolecular Weight
I n applying the viscosity nirthod for asph:iltenes, Xac,k (11) used the equation, .lI = log q,/cK
(6)
\vliercs c \vas c:xpi,essed in w i g h t per ernt. The valuti of K was 3.6 X which was determined from the visiwities ofnaphthali,nc and diphriiyl as solutcbs in benz7c~ni~ and toluene solut,ions. Iris K corlrsponds t o a K,.,, of 3.0 X lo-* in Equation 2 in contrast to 4.48 x lO-~obtaiiicdin t h s picsent \vork. Tlie variatioh is undoubtedly dut. t o its derivation from thi, miilccular in h-,,,, ivciights of naphthalenr and diphenyl in llack'a n-ork, \\-hcrerts the molecular weights o f oil, and rwins were used in this lahoratory.
November 1947
INDUSTRIAL AND ENGINEERING CHEMISTRY 2400
MEAN MOLECULAR WEIGHTS O F ASPHALTENES
T h e cryoscopic method n-as tried for the determination of the molecular weights of asphaltenes. Camphor was used as a solvent because it has a high freezing point constant. The observed molecular weights varied with concentration, as Figure 5 shows. The frcaezing point lowerings of solutions of less than 1 5 concentration !\-ere difficult t o reproduce. For this reason the portion of the molecular weight-concentration curve for concentrations less than I was draivn 7%-ithuncertainty; extrapolation t o zero concentration \vas thus very unreliahle. Although the cryoscopic method can be used to give reliable d a t a for oil and resin fractions. it does not lend itself readily to obtaining molecular weights of asphaltenes. The evaluation of the mean niolecular rveights of aspha1tenr.s by the viscosity method was attempted with benzene as solvcnt by means of Equations 3 and 5 . Table I1 gives typical data, and both eyuatioiis appear equally applicable. Preference is given t,o Equation 5 in vie!!- of statpnients made by Kcnip and Petrrs ( 8 ) , and discussed in a prcvioue section. Benzene has 11ern a convcntional solvent for asphalts and asphalt constituents for inany w a r s in testing Iaboratorirs. Sonic colloidal particlra are present in lx~nzenesolution of asphalt.$, according to Katz and ROII ( 6 ) . Bcnzenc vlutinns of asphaltcrics ehon- Brownian niovcincmt, and some col1oiti:il particles w e unqiii-tionably prcscnt. TT-hrthcr the n u i u l w of tliwc, particles is mfficicnt t o discount t h e viscosity results is uncc'rt:Liii: but clixly+ (3xpcrinic.nts in thiq latioratoi,y wl have slio\vn that only minor poitioils of asp1i:ilt e w s d i ~ p r ~ r in through a col!otlil)n mc~i>il>r:rnc~ at room trmtl nil tht, niol(~c-ul:ir. 1t-t.igIiis
Dropping Cnncn. of K n e i n n t i c Prore-s of P o i n t of Asphnltene-, T.ivosi;y Asphnlt . i s l , h n l t a , G 1100 111. at 77' F . , Maniifncture F. Renzene Cenii+toke 100 134 1c,(i
, . . . . ,
Air-hlomn Steam-reduced
2.00 1 71 1 61 3.08 2 12 2 02 2.11
213 123 144
1G2
n 785 0 770 0 777 0 883 0.793 0 i89 0 790
Ball a n d ring nieth,jd.
25 5 28 3 35 2
'C
\lid-rontinent
i:ulf Coa\t (La.)
T.:
F
G
1Iesiriln
California
Cracked residuum
Flu.; Stenill-reduced Steaiii-rediired Steam-reduced Steam-reduced Stenm-reduced Steam-reduced Stenin-redured Steam-redured .iir-blonn
1 I23
1.273 1.130
1.143 1,147
100 134 lciY ? 13
1442
0 9 23.3 27.7 :33 3
1350 1052 1288 I451 1013
23.5
1no
1210 1170 1170
2i 6
128
30 7 32.4 10 7
162 9 ii
13 0 18 3 23 2
110 1'36
IS 0 22 1
4
!i 1800
-5 u
1600
; .? 1400
ec
5.1200
"
1000
05 10 1 3 2 0 23 Concentration, TT eight I er Cent
Figure .i.Effect of Concentration on RIolecular Weight b> Cr:oscopic 3Iethod
1210 1290 1402
213 Soft 132 169 xtfi
28 pt:iiicc of Equa- Eqilaweights of asphdtc~ncs. tion 5 tion 3 Equation 5 gives niolecnl:ii, iveights for asph 1210 IIM 1299 12GO 700 t o 1800. Tlic otiwrvcd rnolecular w i g h t 1402 1366 1532 1600 with the melting point and sourre of origirial a*phalt ( T a l ~ l e111). lli0 1260 TT-itIi increase in the melting points of a scriw of a-phaltL5from a 1170 126: particular sourcc', tlic percciitagc of :tsphaltencs in 122.5 1120 molecular weights of the asplidtcnes from air-blowi :i-phalts increase n-ith an incrcase in the amount of processing or air l)lowing for a particular source (Figure 6). The molecular weights of the asphalt enes from a series of steam-reduced asphalts do not increase
119 145 106 201 93' 110
44.0
1-14
1.>9 99 1on
1120
Ioio 1031 inin 973 70.3
inno __
80R
IO
a Ball a n d ring.
b Calculated bk Equation 5 ,
1515
Figure
I
-
L 1
,
20 30 40 Asph.tlterie~, P e r C e n t
6. 3Iolecular Weight 2's. isphaltene Coiilerit in Air-Blown Asphalts
I N D U S T R I A L A N D E N G I N E E R I N G CHEMISTRY
1516
with the amount of steam reduction for a particular source (Table 111). Table IV compares the magnitude of molecular weights reported in the literature tTith viscosity molecular weights. The viscosity method gives result6 indicating t h a t the mean molecular weightq of asphaltenes are relatively low. The values correspond with the lower limit of the range of molecular weights rcpotted in the literature. T h e viscosity molecular weights at finite, concentrations are lower than the cryoscopic molrcdar \wight at the same concentrations : Asphaltene Concentration, Keight % 4.7 6.6 8.1
lIolecular Weight Cryoscopic Viscobity (caniphor) !benzene) 1802
2118 ?i13
1433 15.49 I64Y
Thus, on the basis of all comparisons niatlt,, I l i t , vi.cw.it gives Ion. molecular xeight values.
>. tni~thod
Vol. 39, No. 11
SURIkl A R l
The v i s o s i t y method of nieasuring riiolccular weights appears t o be satisfactory for the oil and resin fractions from asphalts. The values for v s p / cand log v 7 / c for oils and resins do not change appreciably with increase in conccnt ration, b u t values for asphaltvncs show a gradual inereax in vhpr'cand log q,!c with increase in concent rat ion. For concentrations of asphaltenes in benzene ltbss thaii 1.0 grain per 100 ml. benzene, differences between [log vi,'cIc and [log v r 'clC = 0 are less than 5yGof [log v,/cIc = 0. Both equations proposed by Staudinger apparently can be applied t o oil and resins from asphalts since [ v a N / c l and c [log vr.!clC for these constituents are linear functions of the cryoscopic niolecular !\.eights. The mean molecular \wights of asphaltcncs viere determined by the viscosity method, using constants based on the molecular w i g h t s of the oil and rcsin fractions. The molecular in time of air blowing \vcxiglits of asphaltenes increasv witli incre (BS indicattd by melting point fiir asphalts from oiic source') of t hi. oiipinal asphalts.
Method 5000-6000 4300-3600 2210-5160 2400 IO60 80,000-140,000 1800 700-1800
Cryoscopic (benzener Cryoiropic (benzene) Crvoiconic (cnnlnhori ~. . , Cryoscopic (benzene) Cryoicopic (naphthalene) Aionornulecular filni Viscosiry of dilute s o h . Viscosity of dilute solti.
Because of the posibility of agglomerated asphaltellea iii l ~ e i i zene solution and the viscosity-iiiolecula~~ w i g h t constant usctl, the validitj- of t h e I o n molecular w i g h t s obtained reniains t o 1x: confirmed. The development of a solvent in xvliich it is ciJi,t:+iii that the asplialtcnes are complc~telyciiswlved ~ o u l deliniitiat~~ oiie disadvantage that can bo clainitd for benzcne solutioii.. lopmenis in elrctron microscope techniquo may ~ I Y J vide required information on t lie actual sizes and iiiol(~ul;w weights of asphaltcw-.
Chemical Characteristics of Banded Ingredients of Coal I
1-THE ~ o r oii k Illinois coak i i i this 1:ihcJratory, c c I t i 4 i ! i w l i h , attentioil has been given t o stutlirs of t lie, physicd :tiid chem-
ical charact,cristirs of the banded ingredients vitraiii, cl:train, durain, and fusain. T h e Stopes clajsificatioii (15) of bniitled ingredients is based on macrovisual appe:trance. It i; of intcrest, therefore, t o coniparc t h t ~ evi.sually selected irigrcdients t o learn .r\-hether each elion-s niorci or less specific cheniical ch:tr.;icteristics. It is also of interest t o learn nlicthcr each ingredient. secured from coals of different ranks, varies ii; c~licmicalcliararteristics as do the source Tvhole coals. Published analyses for the b a ~ i d c ~ itigtdictits l stio\v tr-ielc dit'ferences in chemical composition for ench ingi,etlic~iit. 1'rok)aIil~~ the composition of each ingrctlient, at loasst of Titrain, clarnin, anti undcrgone: durain, is influenced by the t l c g r c ~of tiic't:tiiior~)li~~is