RADIO SPECTRAL LINE STUDIES OF THE INTERSTELLAR MEDIUM

IN THE GALACTIC PLANE

By

ANDREW WILKIN SEACORD, II

A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF

THE UNIVERSITY OF FLORIDA

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE

DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

1975

To My PaAentA

With GwatUude

Louise DeLano Peck Seacotd

and

WiLkin Henry Seacord

ACKNOWLEDGEMENTS

I wish to express my appreciation to Dr. Stephen

Gottesman who introduced me to the intriguing problem of

the galactic ridge and who has provided many comments and

suggestions during the course of my study of it. His en-

couragement and cooperation during the final stages of

this project are greatly appreciated.

The spectra observed as a part of this study were

obtained with the 43 meter telescope at the National

Radio Astronomy Observatory in Green Bank, West Virginia.

I appreciate the opportunity to have used these facilities.

I am grateful to the NRAO staff, particularly the opera-

tors and engineers whose support made the observations suc-

cessful.

I also wish to thank the other members of my doc-

toral committee whose comments and suggestions have been

most helpful. Actually, many people within the Department

of Physics and Astronomy deserve a word of thanks for

their general support, for providing travel funds, and for

making computer time available to me.

Finally, I owe many words of gratitude to my wife,

Penny, who has been patient with me and often provided me

with much-needed words of encouragement and to my daughter,

Becky, who has provided many interesting moments during

the course of my work here.

TABLE OF CONTENTS

Page

ACKNOWLEDGEMENTS...................................... iii

ABSTRACT .................. .................. ......... vii

CHAPTERS

I INTRODUCTION...... ........................... 1

The Galactic Ridge......................... 1

Objective, Method and Scope of

This Study.............................. 4

II INTERSTELLAR MEDIUM ASTROPHYSICS ................ 8

Introduction............................... 8

The Transfer Equation and a General

Solution of It.......................... 8

Thermodynamic Equilibrium.................. 12

The Thermal Continuum............... ...... 14

Radio Recombination Lines.................. 18

Thermodynamic Temperatures and Level

Populations.......................... 19

Population and Depopulation Mechan-

isms of High n-Levels ................ 23

Collision Transition Probalities

and Cross-sections..................... 26

Radiative and Dielectronic Recombina-

tion Coefficients..................... 31

Departure Parameters bn and 6mn......... 34

Radio Recombination Line Absorption

Coefficient .......................... 39

Line Profiles........................... 43

Non-LTE Solution of the Transfer

Equation and Line Enhancement......... 49

The Radiation Field....................... 53

Ionization Rates .......................... 56

The 21 cm Hydrogen Line................. ... 58

Pulsar Dispersion Measure.................. 62

Page

III THE TELESCOPE SYSTEM SENSITIVITY AND THE

OBSERVING SCHEME .... ................ .......... 64

System Sensitivity .............. .......... 65

The Observing Scheme .............. ......... 67

IV THE RADIO RECOMBINATION LINE OBSERVATIONS...... 70

Introduction................................ 70

Observing Scheme and Data Reduction........ 75

The Spectra ................................ 83

V MODELS FOR THE INTERSTELLAR MEDIUM IN THE

DIRECTION 3C391 ............................... 107

3C391: A Summary of Previous Observa-

tions.................... .. ............ 107

The Models................................. 119

The D-Method Models ...................... 119

Isolated Cold Region Models .............. 147

The Hot Spot Models ...................... 156

H II Nebula Models ...................... 175

VI FURTHER CONSIDERATIONS AND APPLICATIONS OF

THE MODELS.................................... 182

Discussion of the Models ................... 182

The Pulsar 1858+03 ........... ............ 200

The 3C396 and COMP Spectra................. 206

Distribution of the Recombination Line

Sources Within the Ridge ................ 210

VII SUGGESTIONS FOR FURTHER RESEARCH ............... 216

APPENDIX. .............. .................... 222

BIBLIOGRAPHY.................................. 226

BIOGRAPHICAL SKETCH..................... ...... 234

Abstract of Dissertation Presented to the Graduate Council

of the University of Florida in Partial Fulfillment of the

Requirements for the Degree of Doctor of Philosophy

RADIO SPECTRAL LINE STUDIES OF THE INTERSTELLAR MEDIUM

IN THE GALACTIC PLANE

By

Andrew Wilkin Seacord, II

August, 1975

Chairman: Stephen T. Gottesman

Major Department: Astronomy

This study develops a model for the source of radio

recombination line emission from the galactic plane. The

model is based mainly on recombination line spectra ob-

served in the direction of the supernova remnant 3C391.

The H157a recombination line was observed and used with the

H92a spectrum observed by Cesarsky and Cesarsky (1973) to

obtain a model consistent with both spectra. Other

criteria--such as the 80 MHz optical depth, the contri-

bution to the thermal continuum temperature, and ionization

rates--were also considered.

Four models were studied: a cold region illumi-

nated by 3C391; an isolated cold region; a group of hot,

dense H II regions; and a low density H II nebula. Of

these, the best model is a small (x85 pc diameter) H II

region with a temperature of 75000 K and an average elec-

-3 -3

tron density of 10 cm

The H157a spectrum in the direction of the pulsar

1858+03 was also observed. It was found that it is prob-

able that one of the recombination line nebulae lies along

the line of sight to the pulsar. In this case the pulsar

would be at a distance of 2.6 kpc in order for its dis-

persion measure to agree with the measured value of 402

-3

pc cm

The large scale distribution of the H II nebulae

was also studied. The results of available radio recom-

bination line spectra surveys from the galactic plane sug-

gest that they exist within an annular region,which ex-

tends from 4 to about 8.7 kpc from the galactic center.

v i i i

CHAPTER I

INTRODUCTION

The Galactic Ridge

As a radio telescope is scanned across the plane

of the Milky Way at some galactic longitude,1 it is seen

to be a source of radio emission. For radio frequencies

of 100 MHz or greater, the intensity increases as the

galactic equator is approached and peaks at the equator

itself. Scans through the galactic plane made at several

different frequencies reveal that the emission has two

components, thermal and nonthermal. At frequencies below

a few GHz, the emission is predominantly nonthermal with

a spectral index between -2.7 and -3.0 (Westerhout, 1958;

Matthews, Pedlar and Davies, 1973). At 85 MHz the radi-

ation is almost completely nonthermal and the intensity

1 II

In this work, the revised galactic latitude, b ,

and longitude, 111, as adopted by the IAU (Blaauw et al.,

1960) will be used throughout. The superscript II, how-

ever, will be dropped from the notation.

distribution across the galactic plane has a total half-in-

tensity width of about 50 at longitude greater than 120

from either side of the galactic center (Hill, Slee, and

Mills,1958). As the frequency increases, the overall in-

tensity decreases and the width of the emission becomes

smaller, approaching a thin ridge at centimeter wavelengths.

At 1390 MHz the half-intensity width is generally about

195. The thickness of the nonthermal emission region in-

creases somewhat near the galactic center. This thin con-

tinuum emission region surrounding the galactic equator

has been named the "galactic ridge."

The galactic ridge has been recently mapped with

an antenna resolution of 11 arc minutes at frequencies of

1414, 2695, and 5000 MHz by Altenhoff et al. (1970).

Matthews, Pedlar, and Davies (1973) have used these maps

to obtain the latitude distribution of the continuum in-

tensities at three longitudes where no resolved sources

are present. They plotted the distribution with the in-

tensity scale at each of the frequencies adjusted so that,

if the emission were entirely nonthermal, the three dis-

tribution curves would be congruent. The plots are con-

gruent at latitudes greater than about 025, but inside this

range, the intensity increases markedly with frequency due

to thermal emission. These curves illustrate the two com-

ponents of the galactic ridge continuum emission.

It is known that hydrogen is concentrated within

a thin region around the galactic equator (cf. Westerhout,

1958; Baker and Burton,1975). The thermal emission comes

from the free-free process between electrons and protons

from ionized hydrogen.

At frequencies below about 30 MHz, a scan cutting

through the ridge shows an intensity profile with two

peaks, one on each side of the equator, The decrease of

intensity along the equator results from absorption of

the strong nonthermal emission by the free-free process

within the thermal emission region. The map by Shain,

Komesaroff, and Higgins (1961) at 19.7 MHz shows the ab-

sorption lane prominently.

The galactic ridge took on an interesting aspect

when Gottesman and Gordon (1970) observed low intensity

H157c radio recombination line spectra in two directions

in which the Altenhoff et al. (1970) maps show no resolved

sources. Gordon, Brown, and Gottesman (1972) found that

at a longitude of 330, the recombination line emission is

limited to a latitude of less than O05 above the equator.

In the region toward the galactic center, recombination

line emission has been observed at a latitude of -0.62

(Lockman and Gordon,1973). Previous to Gottesman and

Gordon's observations, radio recombination lines had only

been observed from known galactic H II regions. They have

not been detected from the interstellar medium in the solar

neighborhood.

A number of workers have studied the galactic

ridge in order to determine the nature of the radio recom-

bination line sources and how they relate to the thermal

continuum and 21 cm spectral line emission. The analysis

of Cesarsky and Cesarsky (1971, 1973b) and Lockman and

Gordon (1973) find that the electron temperature of the

radio recombination line emission regions is close to 20 K.

Other workers derive temperatures ranging from 103 to 104

K (Gottesman and Gordon,1970; Gordon and Gottesman,1971;

Matthews, Pedlar, and Davies,1973; Jackson and Kerr, 1971;

Chaisson,1974a; and Downes and Wilson,1974). Very recently

two papers have appeared,which conclude that these recom-

bination lines originate from low density H II regions

(Jackson and Kerr, 1975; Pankonin,1975).

Objective, Method, and Scope of This Study

The problem to which this study is directed is that

of determining the type of region responsible for the recom-

bination line emission in the galactic ridge. The first

step will be to observe recombination line emission from

regions in the ridge in which there are no H II regions

visible on the Altenhoff et al. (1970) maps and from which

other data have previously been obtained. The recombination

line spectrum will provide one piece of datum, the power

contained in the spectral line. However, there are several

parameters which define the type of region: electron tem-

perature, electron, neutral hydrogen, ionized hydrogen

number densities, size of region, and distribution of the

region, or regions, within the galactic ridge. The latter

may imply a relation between recombination line emission

and other phenomena such as low frequency radio absorption

and galactic structure.

Since the observations made as a part of this study

can only supply one bit of information, the identity of

the recombination line source can be obtained only after

other observations are utilized. Even so, there are still

more parameters whose values are to be found than there

are observed quantities. For that reason, the values of

some parameters must be chosen at the onset of model com-

putation. A type of model is hypothesized and theory or

the conclusions from previous observation suggest what

these values are. The second step will be to obtain, from

one region in the ridge, a model solution which is consis-

tent between two similar data. The ideal case for this

is the region toward the supernova remnant 3C391 from which

Cesarsky and Cesarsky (1973b) have observed an H92a recom-

bination line spectrum. This region has undergone a con-

siderable amount of scrutiny, particularly by those who

have attempted to measure the 80 MHz optical depth (Cas-

well et al., 1971; Dulk and Slee, 1972; Goss, 1972).

The third step in the process will be to assume

that this model applies in the mean sense to the entire

ridge and use the remaining data obtained for this study

to check the model. The method is to calculate a parame-

ter which depends on the model, to see if it agrees with

what is observed. Finally, recombination line surveys

will be combined in such a way as to examine the large

scale structure of the recombination line ridge.

The theoretical basis of this study will be devel-

oped in Chapter II. All equations used in the analysis

will be derived and approximations and assumptions will be

discussed. The third chapter will discuss theory behind

the observing procedure. The method and conditions under

which the observations were made and the display of the

processed spectra will occupy Chapter IV. Chapters II,

III, and IV provide a basis for the subsequent work.

Four types of regions on which to base the models

will be selected. Three of them result from previous obser-

vation of 3C391, a summary of which begins Chapter V. The

Hot Spot model is based on unpublished observation of Alten-

hoff et al. (1973). The remainder of Chapter V develops

the four models. Chapter VI is involved with the selection

of the most probable model and tests this by computing the

distance to the pulsar 1858+03 and the 80 MHz optical depth

expected in the direction of 3C396. Chapter VI ends with

7

an investigation of the large scale structure of ridge

recombination line emission. The last chapter concludes

the study by proposing future projects to study the galac-

tic ridge.

CHAPTER II

INTERSTELLAR MEDIUM ASTROPHYSICS

Introduction

It is the purpose of this chapter to present a

foundation on which this project is based, particularly

the analysis of the spectra. The starting point is the

equation of radiative transfer. The coefficients of ab-

sorption and emission are parameters of the solution to

the transfer equation.

Since this study is based on radio spectral lines,

the line absorption coefficient will be examined in de-

tail. The line profile of a given element depends on the

thermodynamic state, temperature, abundance, electron

density, and motion of the gas creating the line.

The Transfer Equation and a General Solution of It

The transfer equation can be set up by considering

the radiation passing through a region as shown in Figure

II-1. The radiation entering this region has a frequency-

dependent specific intensity, Io(v), which can be both

thermal and nonthermal in origin. The line of sight enters

I() I I io(v)

Q ^ _________--0-r---____ lo

OBSERVER

s=L s=0

-rr T'=O

Figure II-i.

Ray Path Geometry Through a Line-

emitting Region. The region is

shown shaded.

the region at a point where the path length, s, is zero.

The total path length through the region is L. At the

near boundary, where s = L, the specific intensity is

I(v). The change in specific intensity along a differ-

ential path length ds in a homogeneous medium is

dl(v) = KL(v) I(v,s) ds -KT(v) I(v,s) ds + jL(V ) ds + jT(v) ds

+ jN(v) ds. 11-1

The nonthermal emission coefficient is jN(v). In the

interstellar medium the nonthermal absorption processes

are not important. The thermal process, which will

generate or absorb radiation, is the free-free process;

it will be discussed later. The coefficients of ther-

mal continuum emission and absorption are jT(V ,s) andKT(v,s)

respectively. The applicable coefficients for spectral

lines are denoted by the subscript L.

Let j(v) = jT() + jL() and K(v) = KT(V) + L(v),then at any

point,

I(v) = IT() + IN() + IL() = Ic(v) + IL(). Equation II-1 then

becomes

dl(v) j(v) 3N

d ( + + V II -2

The optical depth, T, is defined by

L

t(v) K(v) ds = TC() + TL().

The thermal source function, S(v), is defined by

j(v) jL(v) + jT(v)

S( () = + II-3

-K(0)KL T

Since there is no nonthermal absorption coeffi-

cient and the thermal and nonthermal processes are in-

dependent, there is no nonthermal source function.

A general solution to the transfer equation can

be obtained by multiplying Equation II-2 by the inte-

grating factor exp(--) and integrating along the path

length to get

I(v) = (v ) e- +

~~ (l-e- ) + < (1-e-).~~

The brackets,< >, denote the average over the path

length. The intensity of the nonthermal emission gen-

erated within the line-emitting region is IN = jN>L

and L = T; so, the general solution is

I

I(v) = Ioe I ~~ (l-e ) + (-e-T). 11-4~~

Nonthermal emission decreases in intensity with increas-

ing frequency, in general.

In order to derive a solution for a specific

region, the source function-which depends on the thermo-

dynamic state of the region-must be found. It is im-

portant, therefore, to consider the conditions for thermo-

dynamic equilibrium. This is the subject of the next

section.

Thermodynamic Equilibrium

To be in thermodynamic equilibrium (TE), a gas

must have the following properties.

1. The temperature and pressure throughout the

region are constant and are equal to that of

its surroundings; hence, there will be no

flow of energy or mass through it.

2. The gas will radiate as a black body at that

temperature.

3. Detailed balancing must exist; that is, for

each process occurring in the gas, its inverse

must also occur by an equal amount (e.g.

ionization and recombination).

4. The velocities of the gas particles have a

Maxwellian distribution.

5. The population of any excited state of an

atom, ion, or molecule is determined by the

Boltzmann equation.

6. The relative population of any level of ion-

ization of a species to that of the neutral

species is given by the ionization (Saha)

equation.

The observed spectral line intensity depends on

how well these conditions are met. Radio recombination

line observations of a variety of regions indicate that

the interstellar medium is generally not in local thermo-

dynamic equilibrium (LTE); that is, the conditions of TE

are not met even on a small (local) scale (Hjellming and

Davies, 1970; Andrews, Hjellming, and Churchwell, 1971;

Gordon, 1971; Brown and Balick, 1973). If LTE does not

exist, the transfer equation parameters must be modified

to account for this. In the following work, a super-

script asterisk denotes an LTE quantity. The second con-

dition of LTE dictates that S*(v) = B(v,T), the Planck

function for a temperature T. For radio frequencies,

this simplifies to

B(2,T) kT 2

B(v,T) 2- v2 11-5

c

In the case of non-LTE, a correction parameter is defined

such that

S(v) = A(v) S*(v) = A(v) B(v,T).

II-6

In any steady-state situation (LTE or non-LTE),

the total per-unit volume emission and absorption must

balance. Only in LTE do emission and absorption balance

at each frequency so that, by Kirchhoff's Law, j(v) =

K(v) B(v,T).

The Thermal Continuum

Thermal emission and absorption are the result of

an encounter between a free electron and a positively-

charged ion after which the electron remains free (cf.

Mitchell and Zemansky, 1934, p. 154). From a classical

point of view, an electron moves in the field of an ion

(e.g. H ), is accelerated and absorbs a wave or is decel-

erated and emits a wave. As shown in Figure 11-2, the tra-

jectory of an electron with a given velocity is altered

by an amount determined by the impact parameter, p. For

radio frequencies, p is large enough that the entire

trajectory can be approximated by a straight line.

Oster (1961) has derived an expression for the

thermal absorption coefficient, which includes a quantum

mechanical correction applicable for temperatures in ex-

cess of 5.5 x 105 K. The derivation of K( v ) was ob-

tained by considering emission from a single electron with

specified velocity and impact parameter. Then jT( v ) was

found for an ensemble by integrating first over a suitable

range of p and then over a Maxwellian velocity distri-

bution. Then, KT(v ) = jT(v)/B(v,T) and, for hydrogen,

T(v) = 3.014 x 10-2 T-1.5 v-2 ln(4.955x10-2 T-2.5 -l1) N Z2 N. .

T o 0 e 1 1

11-7

The numerical constants are functions of fundamental con-

stants as shown in the Appendix, Equation A-1. The sum-

mation is over all ionic species present. Z. is the

effective charge number (ZH+ = 1) and Ni is the number

th

denisty for the i species.

Altenhoff et al. (1960) have derived an approx-

imate expression for KT(v),which is

K(x) = 8.235 x 10-2 T-1.35 -2.1 N Z2 N.. 11-8

T e e 1 i I

To bring this approximation into agreement with Oster's

more exact formulation, it is multiplied by a correction

factor a(Te,v). Using it, the optical depth then becomes

TT(V) = (v) ds = 8.235 x 10-2 a(T ,\) T-1.35 -2.1 ET

o1

11-9

ET is the continuum emission measure defined by

L

2

E (N Z2 N.) ds. II-10

T e 1 1

0

-6

The emission measure is expressed in units of pc cm

The factor a(T ,v) has been tabulated by Mezger

and Henderson (1967) for the ranges 100 < v < 105 MHz

and 3 x 103 < T < 15 x 103 K. For these ranges, a(T ,v)

varies between 0.7680 and 0.9971.

The emission measure is often given by

L

E = N2 ds. II-11

T e N

The effect of heavy elements on the size of ET

depends on the fractional ionization of hydrogen; x -

Ne/NH. The atomic species with an ionization potential

less than that of hydrogen (13.6 eV) may be totally

< -3

ionized. If x < 10 the contribution of these elements

to the electron pool will be an appreciable fraction of

the total unless their abundances are less than those

adopted by Spitzer (1968, p. 122). The most abundant

contributor to the electron pool, other than hydrogen,

-4

is carbon which has an abundance NC 4 x 10 NH.

The continuum emission spectrum from a region of

the interstellar medium can provide some information

about the conditions there. Consider Equation 11-3 for

the continuum; that is, IL = 0. Let I = I = 0 and,

assuming LTE, S(v) = B (v,T). Then, using Equation 11-5,

2kT 2 -T

I(v) v (1-e T). 11-12

2

c

When the frequency is high enough that the gas is opti-

cally thin, TT << 1, and

S2kT 2

I(v) = v T.

Since, by Equation 11-8, TT v 21, I(v) a v0) and the

spectrum is nearly flat. For low frequencies such that

2

the gas is optically thick, T, 1.5 and I(v) a The

turnover frequency, vT, is that for which TT = 1, al-

though sometimes the value of 1.5 is used. Setting TT

equal to one in Equation 11-8 and solving for vT gives

vT = 1.893 x 102 a(T ,v)-2 1 ET2.1 T-0.643 11-13

If in some region for which vT can be measured, either

ET or T is already known, the other parameter can be

found. One independent source of information is the in-

tegrated radio recombination line power. Radio

recombination lines will be considered in the next

Section.

Radio Recombination Lines

Radio recombination lines arise from transitions

between levels with large principal quantum numbers;

n 40. They result from the capture recombinationn) of

an electron by an ion into an upper level. The nomen-

clature for such a transition is Xn(An) where X designates

the recombining species; n is the principal quantum num-

ber for the transition which is specified by a for An = 1,

B for An = 2, etc. The predominant element from which

recombination lines are observed is hydrogen; however,

radio recombination lines of helium, carbon, and possibly

heavier elements have been observed (Gordon, 1974;

Chaisson, 1973; Pedlar and Hart, 1974).

The emission frequency, vmn, for the transition

between the mth and nth levels (m > n) is given by the

Rydberg formula (cf. Shore and Menzel, 1968, p. 13)

2 1 1 2 an

v = c R Z ( ) = 2c R Z2 An 11-14

mn x x 2 2 Rx x 3

n m n

R is the Rydberg constant for the species X and Z is

x x

its effective charge number. Radio recombination line

frequencies for hydrogen and helium have been tabulated

by Lilley and Palmer (1968).

The recombination line strength depends on the

number of atoms in the region which are excited to the

upper level of the transition. Since the gas may not be

in LTE, it is necessary to calculate the LTE populations

and the departure parameters,which indicate the degree to

which the gas is out of LTE. These matters will be dis-

cussed in the following subsections.

Thermodynamic Temperatures and Level Populations

Radio recombination line emission regions contain

excited neutral atoms, ions, and electrons. The number

densities (in units of cm-3) for hydrogen are NHo, NH*+

and N respectively. NH is the total hydrogen density.

In LTE there is an equilibrium situation where recombin-

ations are balanced by ionizations. What is desired is

a statistical relation between the kinetic temperature of

the gas, NH+, Ne and Nn, the number density of atoms in

the nth level of excitation.

For a gas in LTE, the population number density of

ions in the nth excited level and the i ionization state

is given by the Boltzmann equation (cf. Aller, 1963, p.

112):

Nn No gn exp(- Xin/kTex). -15

This equation is based on the condition that the

assembly has a Maxwellian velocity distribution. In

Equation 11-15, Noi is the LTE population of the ground

state; gn and go are the statistical weights of the ex-

cited and ground states respectively; Xin is the excitation

energy within the ith state of ionization; and Tex is the

excitation temperature. The statistical weights are the

number of sublevels in the nth level; g = (2J + 1). J

represents the total angular momentum--spin plus orbital--

of the level. Under the conditions of LTE, the excita-

tion temperature, T is equal to the kinetic temperature

of the gas.

The radio recombination lines,which have been

observed involve highly excited neutral atoms, for ex-

ample H I, He I, and C I. So,the desired population is

N expressed in terms of the electron density and the

no

respective number of the singly ionized species. For

the species X,

N (X) = 4.143 x 10-16 N.(X) N Te1 5 exp(- (X)/k T

n 1 e ex 2u7- T) exp(-Ein(X)/k Tex)"

11-16

The details of the derivation of this equation are given

in the Appendix; it results from placing the values of

the fundamental constants into Equation A-5. The par-

tition function, ul(T), is given by Equation A-2. Ein(X)

is the threshold energy which is (Equation A-6)

RADIO FREQUENCY

e

e

X RAY

Figure 11-2.

The Free-Free Process. The elec-

tron, e, approaches the central

ion. The impact parameter, p,

is shown for two cases.

E1n(x)=-(XI-X ) = 157900 (Z 1/n2) k. 11-17

The kinetic energy of a particle with mass M is

EK = 1/2M v2 = 3/2 k TK.

If the gas has more than one constituent, each with a

different mass, the kinetic temperature of each constitu-

ent will be the same although, because of the different

masses, the RMS velocity of each constituent will be

different. The RMS velocity of the component with mass

M is

3k T

2 1/2 3 K 1/2 11-18

If the gas particle velocity distribution should

suddenly become non-Maxwellian, collisions involving a

deflection of 90 will thermalize them to an equilibrium

temperature unless the system is thermodynamically unstable

(Spitzer, 1968, p. 90). In a situation where there is a

nonequipartition of energy, the electrons will have the

largest kinetic temperature because of their low mass.

The energies of the electrons and ions will equalize first

because of the long-range interactions between them.

Following this, there will be elastic collisions between the

ions and neutral atoms. Collisions between electrons and atoms are not

as an effective means of equalizing the kinetic energies

as are collisions between ions and the electrons (Spitzer,

1968, p. 140).

Population and Depopulation Mechanisms of High n-Levels

The high n-levels are populated and depopulated by

several mechanisms which are discussed by Brockelhurst

(1970), Dupree and Goldberg (1970) and McDaniel (1964,

p. 588). The parameters, such as cross-sections, proba-

bilities, and coefficients are discussed in general by

these authors as well as by Spitzer (1968), Kaplan and

Pikelner (1970), and by the workers listed in Table II-1

who have calculated high n-level populations. The mecha-

nisms involved are as follows.

1. Electron recombination followed by a cascade.

2. Dielectric recombination (in atoms with more

than one electron).

3. Excitation and de-excitation induced by

electron collisions.

4. Collisional ionization followed by three-

body recombination leaving an excited atom

X': H + e X+ + e + e and X+ + e + e -

X' + e.

5. Stimulated and spontaneous emission.

6. Fine structure excitation; i.e. redistribution

of angular momentum with a level: nR nt'.

Excitation to high n-levels from the ground level by radi-

ative excitation is not important in H II regions (Dyson,

1969; Dupree and Goldberg, 1970).

To account for the effects of a departure from LTE

with respect to level populations, the departure coeffi-

cient, bn, is employed;

b = N /Nn 11-19

n n n

The departure coefficient is a function of the electron

temperature and density and the intensity of the radiation

field. When b is calculated for a type of region, the

n

radiation temperature is chosen on the basis of specific

environmental conditions of the region and bn is tabu-

lated or plotted as a function of n for specific values of

T and N For example, Dupree (1972a) has done this for

e e

departure coefficients in a cool region. The coefficients

were first calculated for an isolated region where Io(v)

= 0 and then the effects of an adjacent 104 K H II region

on the b values were demonstrated.

n

Statistical equilibrium requires that there are

as many transitions into a level as there are transitions

out of it. The equation of statistical equilibrium for

the nth level is of the form (Seaton, 1964; Dupree, 1972a,

b)

[ n-1 n-1 n+ p' n n-n

N A .+ Z C .+ C .+ ( B + z Bn jl(v)

nL jn n, j 1 nl m=n+1 n,m nn-

'=n 0 =n-P j=ntl m=n+l j= 1

= (N C +N C )+( N B

(Nn-p n-p,n n+p n+p,n m=n+ m m,n

p=l

1 m'

+ j N. B.) ) ( + [ N. A. + N N. (ar + d + )

io ,n j>n+l J ,n e 1 n n i,n

j=no0 j=n+l

II-20

The symbols used in Equation 11-20, which have not yet

been defined, are as follows:

A = Einstein spontaneous emission probability,

m,n

m > n;

B = Einstein absorption probability, m > n;

n,m

B = Einstein stimulated emission probability,

m,n

m > n;

Cj,k = collision probability for a transition

j ,k

between the jt and kth levels;

C n,i = collisional ionization probability from

level n;

C. = three-body recombination probability;

1,n

= rate of radiative recombination into level n;

n

an = rate of dielectronic recombination into level

n ( = 0 for hydrogen);

m' = the maximum level considered in cascades from

higher levels;

p' = the maximum p for bound collisional transitions,

p = |An ;

n = the lowest level of the atom considered for

o

the computation;

and N. = the ion number density.

Collisional Transition Probabilities and Cross-sections

In order to compute the departure coefficients,

means of simplifying Equation 11-20, limiting the number of

equations or limiting the range of the summations should

be considered. At some level n = m', the electron is close

enough to the continuum that the collisional ionization

probability, Cn,i, is much greater than the probability for

a spontaneous transition to lower levels, Am,n. The levels

beyond the m' cut-off level are in thermodynamic equilibrium

and, b = 1 for n > m'.

n

Sejnowski and Hjellming (1969) find that inclusion

of collisional transitions of p = A n > 1 will lower the

value of b by only a few percent from that considering

only transitions of p = 1. Consequently, many workers

limit their calculations to p' = 1.

Collisions involving a change of .- state within

a level, n,R n, + 1, occur frequently. In cold hydrogen

-3-3

gas with N > 10- cm fine structure collisions by pro-

tons as well as electrons,occur more frequently within

levels for which n > 100 than do radiative processes

2

(Pengelly and Seaton, 1964). Because of this, all n

k-states are equally populated. Also, at low temperatures,

fine structure excitation by electron collisions provide

the main cooling mechanism of the gas (Dalgarno and McCray,

1972). By this process, electron kinetic energy is con-

verted to low frequency radiation which leaves the gas,

thus removing energy from the thermal pool.

The total population of level n is divided into

2-states according to

N (2 + 1) N

n,R 2 n

n

(Dupree, 1969). The total population is N = n N

n 2. n,2 "

For hydrogen, the levels are degenerate and b = b

b n is independent of R.

The collisional transition rates, Cjk for

j,k

k = j + p, depend on cross-sections Qj,k by

C N= N . 11-21

j,k e Jk

The average is obtained by integrating over a Maxwellian

velocity distribution.

The bound-bound cross-sections have been computed

by means of one of three approximations or methods:

binary encounter method, impact parameter method, and a

method based on the correspondence principle. The binary

encounter method uses a classical approximation,which is

valid for electron-atom collisions if the time of inter-

action between the incident and bound electrons is short

compared to the orbital period of the bound electron and

the collision takes place in a region smaller than the size

of the atom. This time is sufficiently short when the

energy transferred between the two electrons is large.

With this condition, the energy transferred is assumed to

be equivalent to that transferred between two free elec-

trons. Because of this requirement, the binary encounter

method gives best results for ionizing collisions; the

results are poor for p = 1. In calculating cross-sections,

Brockelhurst (1970) uses the binary encounter method for

p > 6. The application of the theory involved is discussed

at length by Burgess and Percival (1968).

The impact parameter method for electron-atom

collisions uses the assumption that the incident electron

follows a straight-line trajectory past the target atom

with an impact parameter p (cf. Figure 11-2). Seaton

(1962) computes the collision cross-section for a bound

transition (n,R n', j') for both weak and strong cou-

pling. For weak coupling,

Qj,k Pkj(p) 2np dp,

where j = n,z, k = n', V', and po is the cut-off value of

p, which is approximately equal to the smaller of the two

atomic radii of the j and k states. Pkj(p) is the transi-

tion probability as a function of impact parameter.

For strong coupling, Seaton (1962) defines p,

such that Pkj ( 1) = 1/2. For p

mean value of 1/2 so that in this range, Qj,k = 1/2Tpi

For a transition of p = 1, most of Qj,k comes from P>pl

and

j,k = 1/2 p 2 + Pkj(p) 2 p dp.

Seaton (1962) calculates the cross-sections for

the weak- and strong-coupled cases and accepts the smaller

of the two values. Pkj(p) is calculated with the sym-

metrized, time-dependent perturbation theory. Brockelhurst

(1970) claims that the impact parameter method will not

work for p > 1. He assumes the case of strong coupling and

finds that most of Q comes from p < pi for which the per-

turbation theory is not valid.

The classical method only works for large p; the

semi-classical impact parameter method is valid only for

p = 1. The correspondence principle method can be used to

obtain cross-sections for intermediate transitions. All

variations of the correspondence principle state that for

large scale systems,where the uncertainty is much larger

than Planck's constant, h, the result of a quantum mechani-

cal approach is identical to that using a classical ap-

proach. Percival and Richards (1970a) show that the Bohr

correspondence principle can be applied to the energy

transfer of a collisional excitation for the case of a

weak interaction and if the quantum and classical first-

order perturbation theories are both valid. Also, the

collisional probability must be much larger than that of

a radiative transition, a condition which is satisfied for

n > 100. With the the Bohr principle for collisions, the

mean net energy transferred during the collision is equiva-

lent to the energy transferred from independent Fourier

components of order +p and -p of the classical atomic

orbital motion. Pjk(p) is computed from these Fourier

components.

For the case of a strong interaction, Percival and

Richards (1970b) show that the Heisenberg correspondence

principle relates the quantum mechanical matrix elements

with the Fourier components of classical atomic orbital

motion. The results for H II regions are published in a

separate paper (Percival and Richards, 1970c).

For collisional ionization, either the binary en-

counter method (Burgess and Percival, 1968) or the classi-

cal Thompson theory (Dyson, 1969; Seaton, 1964) can be

used to obtain the collision rates. From the Thompson

theory

-15 2 -1/2

C 3.46 x 10 15 n N T

n,i e

The inverse of collisional ionization is three-body re-

combination. Using the equation of detailed balancing

*

N C = N. N C.

n n,i 1 e i,n

Radiative and Dielectronic Recombination Coefficients

The radiative recombination coefficient for cap-

ture into level n is represented by ar. The partial re-

combination coefficient, a(k)' is that for captures into

th -

all levels above and including the k ; n > k. It is

given by a(k) = Za.- Usually k = 2. The total re-

(k) j=k J

combination coefficient a(1), includes radiative cap-

tures into the ground level. Seaton (1959) has derived

formulations for a(1) and a(2) using a symptotic expansion

of the Kramers-Gaunt factor. With a notation similar to

that of Spitzer (1968, p. 117),

(k) = 2.065 x 10-11 Z2 T-112 k (TK). 11-22

The function Ek(T) involves the asymptotic expansion.

Spitzer (1968, p. 117) has tabulated I1 and 2." They are

plotted as Figure 11-3.

Dielectronic recombination occurs in atoms,which

have two excitable electrons. This process is very im-

portant in establishing the high level populations of

carbon atoms and is, in fact, responsible for these levels

being over populated with respect to the LTE populations.

It takes place in two steps, the first of which is a

capture of an electron into a high level, n. The energy,

rather than being radiated, is transferred to a bound

electron,which rises to an excited level; n' << n. This

is an unstable condition and the second step is to dis-

pose of the energy of the lower excited level by means

of a downward transition to a stable level. The energy

difference is either radiated--leaving the atom with a

single, highly excited level--or it is transferred to the

highly excited electron which leaves the atom by auto-

ionization, a process called the Auger effect (McDaniel,

1964, p. 590; Shore and Menzel, 1968, p. 93). The radi-

ative process dominates (Dupree, 1969). The dielectronic

10

Figure 11-3,

2 3 4

0o 10 T 'K o1

C Functions for Calculating the Radiative Recom-

bination Coefficient. C1 is for the total recom-

bination coefficient. These functions are tabu-

lated by Spitzer (1968, p. 117).

recombination mechanism is more effective at populating

highly excited states than is the radiative recombination

mechanism; i.e., ad > a at least for T 104 K and

m i e n n' e

-3

N > 1 cm-3 (Dupree, 1969).

e

Departure Parameters, b and 6

It will be shown later that the spectral line

strengths depend not only on the departure coefficient,

b but--since the transition involves two b 's--it also

n n

depends on the derivative of bn with respect to n. For

that reason, when bn is computed, a parameter related

to the derivative is also computed;

b b

6 m n 11-23

6mn b

m

where m > n.

Departure parameters have been published by sev-

eral workers. Information about this work is summarized

in Table II-1. The first column gives the reference to

the work. The parameters calculated by Dupree (1969)

and Hoang-Binh and Walmsley (1974) are for carbon. The

second and third columns give the range of n over which

the parameters are tabulated or plotted. Columns 4, 5

and 6 give the range parameters used in the calculations:

n m', and p' (cf. Equation 11-20). The seventh column

indicates the method by which the collision cross-sections

Table II-1

Collision Range of T [oK]

eR e

Reference n N n m' p' Model* and N [cm ]

in~~~~~~~~ ~ ~~~ 1 4i__ idA ___ u_________ ____________

Dyson (1969)

Sejnowski & Hjellming

(1969)

Brockelhurst

(1970)

Dupree (1972a)

Dupree (1969)

(Carbon)

80 300

20 240

40 300

50 250

20 300

2 400 4 IP(1)

2 240 20 IP(2)

2 >500 75 (3)

10 600 1 BE(4)

2 600 1 IP(5)

5 x 103
10 < Ne < 105

3 4

5 x 103< Te < 10

10 < Ne < 104

2.5 x 103 < Te < 2 x 104

10 < Ne < 105

20 < Te < 1000

0 < N < 103

Te =104 and 2 x 104

4

1 < N ~ 10

e

Hoang-Binh &

Walmsley (1974)

(Carbon) 80 250

? 500

BE(6)

1 CP

Te = 10 and 100

1 < Ne < 103

*The collisional approximations are BE = Binary Encounter, IP = Impact Parameter, CP = Correspondence

Principle. The references to the cross sections are as follows:

(1) Seaton (1962) (2) Seaton (1962) and Saraph (1964) (3) Brockelhurst (1970): IP for p = 1;

CP for 2 -< p < 6; and BE for p 6 including ionization (4) Flannery (1970) (5) Saraph (1964)

(6) Banks et al. (1973): BE for small p; CP for large p.

were calculated. The reference to the source of the cross-

sections is given in the parentheses following the designa-

tion. Since the departure parameters are functions of the

electron temperature and density, the last column of the

table gives the range of these quantities for which the

parameters were computed.

All, except Hoang-Binh and Walmsley (1974),do not

consider incident radiation in the calculations. Dupree

(1972) does, however, show the effect of an adjacent H II

region on a line-emitting region with a temperature of

20 K. The radiation field from the H II region increases

b slightly for n > 80.

n

The physical significance of departure from LTE

can be illustrated by showing the relation which n, T ,

and N have on b and 6 This is done with Figure 11-4

4 3

in which are drawn curves for T = 10 and 7.5 x 10 K

e

computed by Sejnowski and Hjellming (1969) and curves for

T = 100 K computed by Dupree (1972a). For each temper-

3 -3

ature, there are separate curves for Ne = 10 and 10 cm

As n decreases, collisions become less important

in determining level populations. For small n, radiative

processes dominate and the curves show an asymptotic be-

havior. The heavy portion of the 104 and 7.5 x 103 K

curves in Figure 11-4 is the radiative asymptote. De-

population by stimulated emission is much less frequent

b

0.8

0.7

0.6

0.5

I

10

I 2

10

mn

105

-4

10

-5

10

Figure 11-4.

Departure Parameters for Hydrogen as

a Function of n, Te, and Ne. The curves

with a solid line are for Te = 10i4K and

with a dashed line are for Te = 7.5 x 103'K

(Sejnowski and Hjellming, 1969). The

dotted curve is for Te = 1000K from Du-

pree (1972a).

20 100 180 260

than by spontaneous emission (Gordon, 1974),except when an

intense radiation field is present.

As N decreases, collisions become less frequent

e

and the radiative asymptote extends to larger n. Then, for

hydrogen, there is a decrease in the rate of collisional

excitation of lower n-levels which causes a decrease in

level population with respect to the LTE values as n de-

creases. In the case of carbon, however, dielectronic re-

combination is efficient enough in populating upper levels

that the population will exceed the LTE values for n > 20

% 3 -3

and N < 10 cm- In any case, the populations approach

their LTE values as n - (Dupree, 1969).

It should be noted that for hydrogen, b and 6

n mn

are both positive in the radio regime. For any two levels

m and n, where m > n, b is closer to one than is b In

that sense, the upper level is "overpopulated" with respect

to the lower level. If the slope of the bn curve is large

enough, there is an enhancement of the line intensity over

that obtained from a region in LTE even though the level

population is less than its LTE value. This occurs be-

cause of stimulated emission induced by an exterior source

or the region's own thermal radiation field. An examin-

ation of the radio recombination line absorption coef-

ficien-t and non-LTE solutions of the transfer equation

will show how this happens.

Radio Recombination Line Absorption Coefficient

The rate, R(v) at which a bound-bound absorption

takes place between energy levels n and m (m > n) is equal

to the product of the number of absorbers, N the Einstein

absorption probability, Bnm, and the energy flux, I(v)

dQ/47. The degree of absorption is modulated by the spec-

tral line profile function, L(v). Then

dR

R(v) = Nn L(v) Bnm I( ) 4-

The energy per photon is h nm. The absorption coefficient,

KL (v), is defined such that the energy absorbed along an

increment of path length ds is

dR

dcA -L(V) I(v) d2 = -Nn L(v) Bm I(v) hvnm ds.

However, energy may be emitted by stimulated emission of

the amount

do

de = jl(v) dQ = NM tL(V) Bmn I(v) hymn 4 ds,

and by spontaneous emission by the amount

d s

d S = jS(v) dQ = Nm L(v) A hv ds.

S S m PL mn mn 4r

Smn(v) is the emission line profile, which is usually

identical to the absorption profile. Inserting deA, de ,

and deS in the transfer equation (Equation 11-2) gives

dI) Nm L(v) Amn [Nn L(v) Bn Nm L(v) B mn (

Let PL()) = >L(v) and define a total absorption coef-

ficient, KL(v), by

KL(v) (Nn Bnm B n) L(V) B

B N

-hv B (v) B ( mn Nm)

4 () Bnm B nm N

Using gn B = g Bmn (Mihalas, 1970, p. 83), the Boltzmann

n nm m mn

equation, and the departure coefficients, the absorption

coefficient becomes

hv b-24

KLU) L(v) Bnm bn N[ exp(-hv/k Tex)] 11-24

For LTE, b = b = 1 and the LTE absorption coefficient is

m n

K (V) = () B Nm [1 exp(-hv/k Tx)].

L ( f) = nm n ex

11-25

The Einstein probability can be written in terms

of the oscillator strength, fnm

2 2

B = 4 T e f I 1-26

nm hv Mc nm

In the above equation, e is the electron's charge; M is

its mass. Oscillator strengths for hydrogen radio re-

combination lines have been computed by Goldwire (1968)

and Menzel (1969).

Substituting Equation 11-26 into Equation 11-25

and assuming that hv << kTe, the absorption equation be-

ex

comes

2

h v T e e

ex

N can be obtained from the Saha-Boltzmann equation

n

(Equation A-5) and mn can be obtained from an approxi-

2 3

mation to the Rydberg equation; vmn = 2c RH Z An/n

Using these and the appropriate constants (RH = 1.097

x 105, Z = 1, u(T) = 1, and g = 2n2), the LTE absorp-

tion coefficient becomes

KL( -) = 1.071 x 107 (v) f N. N T-25 exp( /k T ) 11-27

where n nm 1 e e n ex

where cn /k = 157890/n2. The LTE optical depth is

n

TL(V) = 1.071 x 107 () fnm T-25 exp(En/k Te) EL. 11-28

EL is the line emission measure defined by

L

JEL N N ds

o

and N. is the number density of the recombining ion.

Following the notation of Dupree and Goldberg

(1970), mn is defined by

b

Bm E [1 m exp(-hv/k Te)] / [1 exp(-h-/k T )]. 11-29

mn b ex ex

n

Then KL(v) = b mn K L() and TL(v) = bn mn TL(V).

In most cases for radio recombination lines, hv/

k Tex is small enough that the exponentials involving them

are equal to 1 hv/k T Then,

k T bm he b k Tex b b n

5mn kI [1 ( ( )] = [1- ( b )].

n ex n m

Using mn, which was defined in Equation 11-23,

kT

Smn ( 1 ) (1 6 ) 11-30

mn )1-6 hv mn

mn

If Ab = b b is small, Ab db and

m n

= db d(ln b) An

mn bm dn m=n

which is a notation often used.

Gordon (1971) and Hjellming and Gordon (1971)

define a different B than that defined in Equation II-29;

kT kT

8 1 ex 6 ) = 1 ex d In b

Bmn hv 6mn hv dn A

This is given for reference only since the definition of

Bmn in Equation 11-29 will be used in the present work.

Line Profiles

There are three line profile functions,which are

relevant to radio recombination line analysis. They are

the Gaussian, Lorentzian, and the convolution of these

two, the Voigt function. The Gaussian function arises

from the Doppler shift resulting from thermal, turbulent,

expansion, and contraction motions of the line emitter

relative to the observer. A Lorentzian profile for a

radio recombination line results from pressure broadening.

Natural broadening, or radiation damping, also produces

a Lorentzian profile, but this is such a small effect that

it is impossible to detect in the observed line profiles.

If pressure broadening occurs, the effect would be blended

with the Doppler broadening and a Voigt profile would result.

Doppler broadening. Mainly, thermal broadening

will be considered. The thermal motion of particles in a

gas with a kinetic temperature TK has a Maxwellian velocity

distribution; so, the number of particles, out of a total

of N, which have a radial velocity between v and v + dv

is

d N(v) exp[-(v-v ) /v ] dv.

In this equation, v is the most probable velocity and

v is the velocity of the ensemble as a unit. v =

(2k TK/M)1/2 where M is the mass of the particles. The

intensity of the spectral line at a frequency Vl is pro-

portional to the number of emitters Doppler-shifted to

that frequency. The line profile is a Gaussian;

1(v) = exp[-(v-v ) /v2] dv. 11-31

The profile function is a probability distribution;

S$(v) dv = 1.

The power spectrum is usually displayed in terms

of antenna temperature as a function of velocity. Velo-

city is physically related to the emitter and the line

half-power width is then invariant to the line rest

frequency. The peak line antenna temperature, T'(c),

is that for which v = v The temperature profile is

T'(v) = T'(c) exp[-(v-v ) /v ]. 11-32

c 0

The integrated line power, which is the area under

the profile, is

P = I T'(v) dv = T'(c) vo F 11-33

Since v is not known, it is desired to substitute it with

a measurable quantity. When the line power drops to one

half of its peak intensity, v = v1 and

P P 2/2

T'(v) = 1/2 P- Pexp[-(v-v) /v2]

vo \ T 0 Vo TTc

The Doppler half-width is defined by AvD = 2(vl-vc); then

8k T

AvD = 2v 01n = (- D -M n 2) 11-34

TD is the Doppler temperature which is the effective

temperature of the gas if the line width were the result

of thermal motion only.

The integrated line power in units of K km/s is

P(K km/s) = (4 )1/2 T'(c)AVD.

However,P is usually given in units of "K kHz; so,

P(oK kHz) ( n 1 2)1 ( ) T'(c) Av

4 ln 2 c D

= 1.0645 ( ) T'(c) Av II1 -35

In Equation 11-35, v is the rest frequency measured in

kHz; the velocity of light, c, and the Doppler width are

both measured in km/s. Likewise, the integrated optical

depth, TL, is

TL = L( ) dv = 1.0645 T(vc )( ) Av 1-36

The Doppler temperature is given by

TD k n (AV )2 11-37

Doppler broadening includes the effects of thermal motion

by an amount AvT, turbulence of a scale smaller than the

antenna beam (Aturb), and expansion or contraction of the

emitting region (AVx). These velocity components combine

to give a total Doppler width of

Av = (Av2 + urb +Av /2 11-38

0 T turb

Pressure broadening and the Voigt profile. When

the electron density is high enough that the time be-

tween collisions is shorter than the life-time of the

excited atomic level, the energy distribution of the

radiative transition is altered. The wavetrain is

abruptly terminated by the passing electron. The line

profile emitted by the assembly of atoms is a Lorentz

function (Kuhn, 1962, p. 392). Peach (1972) has shown,

however, that the profile is not an exact Lorentz func-

tion, but it is close enough for an accuracy of about 10

percent. The main effect of the Lorentzian is to redis-

tribute the emitted energy to the wings of the line.

Griem (1967) has shown that proton-atom collisions are

not important.

Brockelhurst and Seaton (1972) have determined

that the pressure broadened half-power width for a m n

transition is

n 4.4 10 0 1 N

vP = 9.4 ( 0) ( ) e

e

Pressure broadening will occur in addition to Doppler

broadening. If the gas has an electron temperature T

and a Doppler temperature TD (Equation 11-37), the ratio

of pressure to Doppler broadening is, by Brockelhurst

and Seaton (1972),

0.12 7.4 Ne 2x104 1/2 04 0.1-

- 0.142 ( I ) 4) ) 11-39

AVD 10 D e

To obtain appreciable pressure broadening of radio re-

combination lines, TD > 2 x 104 and Te > 10 and Ne

> 104. So far, the only possible radio recombination line

sources known to have such parameters are dense knots in

emission nebulae such as M42, W51, and W3 (Hjellming and

Davies, 1970) or dense H II regions associated with type

I OH masers (Habing et al., 1974). No pressure broadening

has been seen, however (Gordon, 1974).

When Doppler and pressure broadening occur to-

gether, the resulting profile is the convolution of a

Lorentz with a Gaussian which is a Voigt profile, pV(v).

The convolution integral cannot be evaluated exactly. So,

the function is expressed in the form

(V) = A 1 H(a,w). II-40

where H(a,w) is the Voigt function

2

a _-y

H(a,.) a e- 2 dy

Sf (~-y)2+a2

and

v v a R

= y and a --

CY o U

The frequency dispersion, o, is given by a = vD 2(ln 2)1/2

= v ( -) (In 2)1/2. Posener (1959) has tabulated the

D c

Voigt function for values of astronomical interest.

Non-LTE Solutions of the Transfer Equation and Line

Enhancement

Two solutions of the transfer equation will be

developed here for use in the analysis of the spectra.

They will be derived starting with Equation 11-4. Since

the frequency of the spectral line is high, the Rayleigh-

Jeans approximation to the Planck function can be used to

convert specific intensity to brightness temperature.

From Equation 11-4, the brightness temperature is

T N -T

TB() = TL(v) + TC() = A(v) Te (l-eT) + (l-e-) Te

11-41

where T= TL(V) + TT(V).

Since, from Equation 11-6, A(v) = S(V)/B(v,T )

and jL(v) = b KL B(v,T ),

mT e

LT + b KL 1 + bm (L/KT)

A(v) = m II-42

T +n mn L 1 + bn mn (KL/T)

Optically thin H I region adjacent to a strong

source of radiation. Assuming that KL>> KT at the line

center (i.e. atv = vc),

b (L/T) bm

A(v) m = m

bn mn (KL/ K) bn mn

Since the region is optically thin, Equation 11-41 re-

duces to

b

TB(v) = Te ( m TT L) + TN + T (1-i T-L)'

When the frequency is off the line (i.e.v>>vc) only the

continuum is observed;

b

TB(vV)= TC( = bn Te T + TN + T (1-TT)

Subtracting this equation from the preceding one gives only

the line;

b

TL() = T () TC() = T T T

n mnne Lo

Since L = b Bmn T'

TL (v) = L(V) (bTe b Bm To). 11-43

This is the line brightness temperature from an optically

thin H I region for which KL >> KT. The second term in

Equation 11-43 will produce line enhancement if Bmn is

negative. To convert the brightness temperature to an-

tenna temperature, the right side of the equation must

be multiplied by the beam dilution factor, ,B' and the

main antenna beam efficiency, B. These are defined by

PB QS/ M and B M/A, where QS is the solid angle of

the line-emitting region; QM is the solid angle of the

antenna main beam; and qA is the total solid angle of

the antenna beam. The antenna temperatures are denoted

with a prime; so, Equation 11-43 becomes

TL( ) = PB B TK(v) (bm Te bn mn To). 11-44

Isolated region. This solution of the transfer

equation will be for the case of an isolated emitting

region for which the incident radiation is weak and can

be ignored. Then T = TN = 0 and Equation 11-41 reduces

to

TB(v) = A(v) Te [1-e ]T

Off the line, A(v) will be unity; so, the continuum tem-

perature is

TC(v) = T (1-e TT)

Taking A(v) from Equation 11-42 and converting TB(v)

to antenna temperature, T(v) = TB(v) TC() and

ST + b L -( T -T

TL( = B B Te b [l-e (1-e .

T + bn nm TL

11-45

Line enhancement. Equations 11-44 and -45 will

be used for the analysis of this study's observations.

The equations indicate how non-LTE conditions affect the

radio recombination line temperature. For LTE b = b

= n = 1; 6 = 0; and Equation 11-44 becomes TL())

= L (T -T ).

eL e

Figure 11-4 shows that as N decreases, collisions

become less frequent; so, the level populations and b

decreases. As the departure from LTE increases, 6mn be-

comes larger. When 6mn > (hv/kT), B is negative and

amplification of the line takes place. This process is

often referred to as the "partial maser" process. The

populations of the transition levels are not actually

inverted as is apparently true for, say, 1120 masers

(Sullivan, 1973).

The line enhancement can be seen from Equations II-

44 and -45 in which TL = bn mn TL is negative, Usually,

L << T because the path length involved with 'T

is much larger than that involved in

TL. Also, b is always positive. Therefore, if b

mn TL/T TI < 1, A(v) increases and the total absorption,

TL + TT' decreases.

The line enhancement will be lessened somewhat

because b < 1. However, for a decrease in electron

m

S3 -3

density from 10 to 10 cm 6 increases by more than a

mn

factor of ten while b decreases by about 14 percent.

The net result is a substantial increase in line power.

Equation 11-44 also indicates the possibility

that the recombination lines may be seen in absorption

against a strong radio source if the relative magnitudes

of b and 6 are such to make B positive. However even

n mn mn

for the case of a dense H II region (EL % 107 pc cm 6)

seen through a tenuous H II region, Brockelhurst and

Seaton (1972) have shown graphically that absorption

cannot be greater than 8 percent. So far, radio re-

combination lines have not been seen in absorption.

The Radiation Field

The radiation field intensity entering a recom-

bination line-emission region, and used in Equation 11-20

to determine the level populations,is computed from I (v)

= W(v) B(v,T). Usually the radiation source is a nearby

nebula whose blackbody temperature is 104 K. The dilu-

tion factor, W(v), is a product of three factors;

W(v) = WG WA(v) A(v). 11-46

WA(v) represents absorption of the radiation by the in-

tervening medium between the source and the line emission

region. If T is the intervening optical depth, WA(v)

= e If the nebula is adjacent to the line-emission

region, WA(v) = 1.

A(v) represents the departure of the radiation

source from a blackbody. It is defined in Equation 11-6.

The geometrical dilution factor, WG, can be

found as follows. Consider a radiation source which is

roughly spherical in shape with a diameter ds and located

at a distance rR from the line-emitter. The geometry

is shown in Figure II-5a. The line-emitting region sees

the radiation source as a disc of angular diameter

2

and filling a solid angle OR = R 2 /4. Then

2 2 R

Ti d2/4 d 2

S s s R 11-47

G = 2 2 4 -47

4 i rR 16 rR

If the two regions are contiguous, rR = 1/2 ds

and WG = 1/4. If, however, the radiation source is much

larger than is the line emitter, Figure II-5b applies and

WG = 1/2. Finally, if the source of radiation surrounds

the line emitter or if the line is emitted by a dense

H II region, WG = 1.

.R .

a

bR

b

Figure 11-5.

Determination of the Geometrical Dilution

Factor, WG. Part a shows the line emit-

ting region and the continuum source to

be about the same size and rR > ds. Part

b shows the line emitting region adjacent

to a large continuum source.

LINE

EMISSION

REGION

CONTINUUM

SOURCE

A numerical example will illustrate the case

where rR > ds/2. Consider the radiation of temperature

TR entering a small line emitting region located 100 pc

from an H II region, such as NGC 7000 (the North American

Nebula), for which T = 104 K and r = 25 pc (Kaplan and

e s

Pikelner, 1970, p. 152). For this case A(v) = 1. Assume

that WA(v) = 1 also. Then, using Equation 11-47, it is

found that TR = 131 K. This indicates that, unless the

line-emitting region is directly adjacent to a large H II

region, the line-emitting region can be considered as

being isolated. In the case where the line emitter is

itself an H II region (T = 104 K), Dyson (1969) shows

that the effect of the free-free continuum on the departure

parameters is very small. Dupree (1972) and Hoang-Binh

and Walmsey (1974) demonstrate the effect of a large H II

region adjacent to the line-emitting region.

Ionization Rates

Knowing the radiative recombination coefficient,

a(T), the required rate of ionization per atom of species

X, (X, can be determined from the condition of detailed

balance. For hydrogen,

NH (H = N NHt a(T).

II-48

The radiative recombination coefficient may be obtained

from Equation 11-22. Habing (1970) has a formula for

computing CH which uses the assumption that a(T) T T0.65

= 1.27 x 10-15 2.1 T-0.3 TT(T)/ 21(v) dv.

11-49

In this formula, TT(VT) is the continuum optical depth

at a frequency vT and 21 (v) is the 21 cm line optical

depth discussed in the next section.

Ionization rates can be calculated theoretically

from a given source of ionization such as low energy cos-

mic rays (Dalgarno and McCray, 1972) and OB stars. These

rates can be compared to the ionization rate required by

a model for a line emitting region.

As an example the ionization rate within the

Stromgren sphere surrounding an OB star will be considered.

Let S be the total number of ultraviolet photons with

u

energy greater than hvL emitted from the central star.

vL is the frequency of the Lyman limit. Also let rS be

the Stromgren radius which is the radial distance from the

star at which the ionizing radiation flux is zero. Then

S = 4 r3 N NH+ a(T)

Su =3 = S e

(Spitzer, 1968, p. 116). From Equation 11-48, a(T)

= NH H/Ne NH+ and the ionization rate of an OB star is

3S

u TI-50

H H

4 rS NH

The ultraviolet flux, S from stars of spectral class 05

through B1 are tabulated by Spitzer (1968, p. 117).

The 21 cm Hydrogen Line

The well known 21 cm hydrogen line is emitted by a

magnetic dipole transition between the hyperfine levels

of the ground state. This is the result of the difference

of energy levels between the proton and electron spins

being parallel (upper state, F = 1) and antiparallel

(lower state, F = 0). The frequency of the transition

is 1420.405 MHz. The upper state is metastable; the

spontaneous emission probability is A10 = 2.85 x 10 15

-I

sec (Kaplan and Pikelner, 1970, p. 410). Usually the

metastable state is depopulated by collisional excitation

more frequently than it is by the 21 cm radiative transi-

tion. The F = 1 state is populated by atom-atom col-

lisions (Kaplan and Pikelner, 1970, p. 420).

Designating the lower and upper states by 0 and

1 respectively, the absorption coefficient is

KhN) (lv) [NO B N B]

10 (v) [N 0 01 1 10 ]

The parameters are analagous to those of the recombination

line. Since collisions play an important part in popu-

lating the upper state, it will be assumed for the moment

that the populations are LTE values as determined by the

Boltzmann equation. The statistical weights of the states

are gl = 3 and g0 = 1. Then, since g0 B01 g B10 and

hv10/kT << 1,

hv hv

10 Tf0 12 0 01

KlO(V) 40() k- s-No B

T is the excitation temperature which, for the 21 cm

transition, is called the spin temperature because the

transition results from a change of electron spin orien-

tation. The absorption coefficient is expressed in

terms of the spontaneous emission coefficient, A10;

2h v0

10 2 B 10

Then

3N h c

K 0 A (N) 11-51

10 k Ts A10 (10 ) II-51

The absorption coefficient is used to obtain the

neutral hydrogen column density along a line of sight.

For a cold gas, the total neutral hydrogen density, NHO,

is NO + N1. From the Boltzmann equation, the number den-

sity at some point s in the line of sight is

NH(s) = No(s) [1+3(1-hv10/k Ts)] = 4No(s),

since h 10/k = 0.07. Substituting in values for the

constants and integrating over the line profile and along

the line of sight gives

L L

L 1(v,s) ds dv T21(v) dv = 2.58 x 1015 H ds,

v o

II-52

where the path length, L, is in cm and v10 is in Hz.

If the line profile is Gaussian with a half-power

width vD and a peak optical depth 21, Equation 11-36 can

be used to obtain

a s-- ds = 4.126 x 101421 Av 11-53

It may be necessary to determine whether or not the

spin and gas kinetic temperatures are equal. Starting

with the equation of detailed balance,

N1 NH Co1 (TK) + N1 A10 + N1 B10 I(10TTR)

= NO NH C01(TK) + NO B 1 I( 10,TR).

C01(TK) and C o1 (TK) are the collisional excitation and de-

excitation probabilities, respectively, whose formulation

is given by Kaplan and Pikelner (1970, p. 420). Solving

the detailed balance equation for Ts in terms of TK and

the radiation temperature, TR, gives

557.3 T0.5 N + T

T -= K 11-54

557.3 T -0 5 N + 1

K H

T = TK if the hydrogen density is large enough that

557.3 T0.5 N >> TR For H I regions, TK > 64 K and

K H R K

N > 1 cm-3 is not unreasonable (Baker and Burton, 1975);

e

so 557.3 T0.5 N > 4.4 x 103 K which is much larger

K H

than one would expect TR to be. Therefore, unless NH

-3

< 1 cm Ts = TK.

Absorption spectra of the 21 cm line have been

obtained by three techniques. Hagen, Lilley, and McClain

(1955) obtained absorption spectra in the direction of

discrete sources by averaging the spectra from off-source

beams and subtracting this from the on-source beam spectrum.

The accuracy to which this method yields the true absorp-

tion spectrum depends on how uniform the hydrogen gas is

over the grid of beams.

Radhakrishnan et al. (1972a) and Hughes, Thompson,

and Colvin (1971) have observed absorption spectra in the

direction of radio sources (e.g. 3C391) with two-element

interferometers. The front-ends and synchronus detectors

are switched in such a way that the system is insensitive

to all but the absorption spectrum in the direction of the

source.

A third method has been used to obtain absorption

spectra in the direction of pulsars. Some of this work has

been done by Gordon et al. (1969), Manchester et al (1969),

Gordon and Gordon (1970), Guelin et al. (1969), and

Encrenaz and Guelin (1970). The pulsar is observed with

a gated, spectral line receiver which alternately stores

and integrates the spectrum with the pulsar on and with

it off, At the end of the integration period, the pulsar-

off spectrum is subtracted from the pulsar-on spectrum.

The difference of the two is the absorption spectrum. The

pulsar is a particularly useful beacon because its dis-

persion measure provides a value of the integrated electron

density along the line of sight. Therefore, two pieces of

information about the intervening interstellar medium are

known.

Pulsar Dispersion Measure

The arrival time of the pulsed radio emission from

pulsars is frequency dependent. The pulse travels at the

group velocity. The group index of refraction is a func-

tion of the electron density of the medium through which

the pulse travels. The difference of arrival time, t1 t2

of the respective frequencies v1 and v2 is, then, the in-

tegrated electron density along the path length L. This is

the dispersion measure, DM, which is calculated from the

equation (cf. Terzian, 1972)

L

OM N ds = 2.4 x 10-4 (t -t2) ( 2- 2) 1

II-55

From this equation, the unit of DM is pc cm-3

From this equation, the unit of DM is pc cm

CHAPTER III

THE TELESCOPE SENSITIVITY AND

THE OBSERVING SCHEME

The observations, which will be discussed in the

next chapter, were made in such a way as to make the most

efficient use of the telescope. This chapter will discuss

the basis for the observational scheme. The important fac-

tor was the system sensitivity which is affected by the

microwave frontend of the receiver and the one-bit auto-

correlation spectrometer.

The general theory involved with the operation of

a one-bit autocorrelation spectrometer for astronomical use

was discussed in detail by Weinreb (1963). The operation

of the instrument used for the observations is discussed

in two reports. The IF filter system, which precedes the

correlator and determines the width of the spectral window,

is described by Mauzy (1968). The operation of the entire

system, with emphasis on the digital autocorrelator, is

described by Shalloway et al. (1968). A discussion of

system sensitivity for the different methods of observing

radio spectra is presented by Ball (1972).

System Sensitivity

The nominal system sensitivity is the RMS noise

temperature, AT, given by

yT

AT = ys III-1

/Af t.

The parameters are as follows:

y = a constant which depends on the system and the

method of its operation;

T = the system noise temperature;

sys

Af = the frequency resolution bandwith; and

t. = the integration time.

T is the combination of the noise temperatures

sys

received by the antenna as well as the power generated by

the antenna-transmission line system and the receiver.

The total bandwith, B, observed by the spectrom-

eter is divided up into N channels. Both B and N are

selectable. Each channel has the effect of measuring the

spectral power within a filter whose bandpass is of the

form (sin v)/v having a half-power width of

Af = 1.21 B III-2

(Shalloway et al., 1968). The spacing between channel

centers is B/N.

The factor y in Equation III-1 is greater than one

and is the degradation of the system sensitivity due to the

operation of the autocorrelator. It is a one-bit digital

machine which computes the autocorrelation function, p(nAt),

of the undetected IF power; nAt is the time delay at the

nth channel. The first step in obtaining the autocorrela-

tion function is to clip the signal, a random Gaussian volt-

age, v(t). The clipper generates a function, y(t); y(t) =

+Y if v(t) > 0 and y(t) = -Y if v(t) < 0. The function

y(t) is then sampled at a rate equal to 2B = I/At. The out-

put of the sampler is fed into an N-bit shift register and

N one-bit multipliers. The shift register is updated at

the time intervals of At. The output of each of the multi-

pliers updates a counter. At the end of the integration

time, each counter contains the sum which represents one

point of the autocorrelation function. The function is in

the form of an array which is transferred to the on-line

computer which computes the power spectrum by taking the

Fourier transform of it.

The effect of using one-bit quantization is to

increase AT above that which would result from a many-bit

quantization (lleinreb, 1963, p. 32). This is accounted

for in the constant y of Equation III-1. This constant

is actually a product of two factors; y = aB. The ef-

fect of one-bit quantization is that B = 1.39 (Weinreb,

1963, p. 49).1 The value of the other factor depends on

the weighting function used in computing the correlation

function. A uniform weighting function was used for which

a = 1.099 (Weinreb, 1963, p. 22). Therefore, v = 1.528

and Equation III-1 becomes

1.528 T

AT = s8 TSys 111-3

/Af t.

The Observing Scheme

There are several observing schemes which can be

used to observe spectral lines (cf. Ball, 1972). The ob-

serving scheme must provide a reference spectrum which is

used to remove the effects of the receiver passband from

the desired spectra.

The "Total Power" scheme can provide an efficient

means of obtaining the reference spectrum. Before observ-

ing a source on the program, a spectrum is obtained while

the antenna is pointed in a direction close to the program

1Burns and Yao (1969) have determined that the clip-

ping factor, B, is a function of the sampling rate. For

a sampling rate of 2B, 0 = T/2 = 1.571. If this is cor-

rect, y = 1.726 and the RMS noise of the data discussed in

Chapter IV should be multiplied by a factor of 1.13. How-

ever, the data reduction routines continue to use B = 1.39

(Cram, 1973, p. 133).

source and where the spectral line is not detected. This

spectrum is the receiver passband function. The nomen-

clature of the Total Power method is to designate the ref-

erence spectrum by "OFF" and a program spectrum by "ON."

If the receiver passband remains unchanged while the OFF

and ON spectra are observed, subtracting the OFF spectrum

from the ON spectrum removes the effect of the passband

from the desired spectrum. The efficiency of this observ-

ing scheme is realized if one OFF spectrum can be used with

more than one ON spectrum if there are several sources in

the observing program.

Let the observing time on an OFF source be tF and

that on an ON be tN. If an OFF source is followed by n

ON sources, each of which is observed for a time tN, the

total observing time, to, will be to = tF + n tN. The

times tF and tN and the ratio n must be chosen so that

the receiver passband remains unchanged during the time

to. The most efficient use of observing time for a given

number of program sources will be obtained if the ratio

a = tF/tN is chosen to minimize AT.

When an OFF spectrum with an RMS noise (AT)F is

subtracted from an ON spectrum with an RMS noise (AT)N the

RMS noise of the resulting spectrum will be

2 2 1/2

T = [(AT) + (AT) 2

N F

For spectral line observations, Tys is not expected to

change more than a few percent between the OFF and ON po-

sitions; then an average system temperature, Tave, can be

used to get

yT

ave

AT- ve

SAf

+

Referring to Equation III-1, the integration time is

tN tF

t. N F 111-4

1 tN tF

Since tF = a tN and to = (a + n)tN,

T

AT = ave (a + 1

/Af to

The value of a is determined when

+ n + n/a)

T is a minimum, for which

d(AT) .

da

Then (1-n/a2)(a+l+n+n/a)-1/2 = 0. The only reasonable

solution is a = / Therefore, the most efficient Total

Power observing pattern exists for

tF = /n tN

III-5

CHAPTER IV

THE RADIO RECOMBINATION LINE OBSERVATIONS

Introduction

The H157a radio recombination line spectra for

this study were obtained between 18 and 24 April, 1973

with the 43 meter radio telescope of the National Radio

Astronomy Observatory (NRAO) located in Green Bank,

West Virginia. The rest frequency of the line is 1.683200

GHz. The H157a line was chosen for two reasons. It is

conveniently located within the passband of the NRAO 10

cm cooled receiver which has good sensitivity and a rep-

utation of being very stable. Also, the frequency of this

line is such that the line strength is relatively strong.

The total power observing scheme was used to obtain the

spectra.

The sources used for the observations are the two

supernova remnants (SNR) 3C391 and 3C396 and the pulsar

1858+03. These were chosen on the basis of the objectives

discussed in Chapter I; they are located in the galactic

1NRAO is operated by Associated Universities, Inc.

under contact with the National Science Foundation.

ridge and independent data from the intervening inter-

stellar medium are available. A radio galaxy, 3C295 (Gold-

stein 1962), was also observed, but its galactic lati-

tude is 610 and no radio recombination line was observed

in its direction. The purpose of observing it was to dem-

onstrate that the recombination lines came from the ridge

gas and were not affected by instrumental effects.

Along with 3C391 and 3C396, a region adjacent to

each of them was also observed. These will be referred

to by the designation COMP. The COMP spectra were observed

for two reasons. The primary reason was to be able to

subtract out emission from the region immediately surround-

ing each of the SNR's and, thereby, make these ob-

servations compatible with those made with narrow beams.

The second reason for obtaining the COMP spectra was to

determine the angular scale over which the medium changes.

Table IV-1 lists the center position of each beam.

The right ascension and declination (1950.0 epoch) and

the galactic longitude and latitude are given for each of

the regions studied. The OFF beams are the reference

beams used for the Total Power observing scheme. Their

selection and use will be discussed in the following sec-

tion.

Figure IV-1 shows the continuumradiation environ-

ment of the regions observed. The continuum maps are adapted

from the 1414 MHz maps of Altenhoff et al. (1970). The

Table IV-1

Antenna Beam Positions

a(1950)

1 46m 47.s8

1 46m 01.s5

1 01m 35.s7

1 59m 59.s2

1 30m 0. S

1 58m 40.S0

1 12m O. O

6(1950)

S58 48."0

S 53' 14."1

S22' 30."0

' 16' 49."7

S30' 0."0

S27' 2."0

S00' 0.'0

beam for 3C391, 3C396, and their COMP beams

beam for the pulsar, 1858+03

Beam

3C391

COMP

3C391

3C396

COMP

3C396

OFFa

1858+03

OFFb

lII

11

31. 9

32. 1

39.02

38.9

35. 0

37.2

37. 4

bII

b1

0.0

-0. 2

-0.3

0.0

+5.00

-0.6

-4.02

athe OFF

bhe OFF

the OFF

Figure IV-1. The Continuum Radiation Environment of the Program Regions.

The size of the circles represents the antenna beam half-

power width, 18 arc minutes. The 1414 GHz continuum maps

are from Altenhoff et al. (1970). The counter is 1.74 K

of brightness temperature.

.1

size of the circle at each region is about equal to the half-

power width of the antenna beam, 18 arc minutes at the

H157a frequency.

Observing Scheme and Data Reduction

The receiver front-end was mounted at the prime

focus of the 43 meter paraboloid. A scaler horn was lo-

cated at the focal point of the reflector. The signals

of linear orthogonal polarizations were separated behind

the horn and fed into separate and identical receivers.

The box containing the front-end equipment was rotated so

that the E-plane of the horn was aligned normal to the

celestial equater. Each receiver was load-switched and

contained a two-stage parametric amplifier. The first

stage of the parametric amplifier and the switched load

were refrigerated.

The local oscillator frequency was generated by

a frequency synthesizer followed by a 6x frequency multi-

plier. The intermediate frequency (IF) sent to the con-

trol room from the front-end box was 150 MHz. The synthe-

sizer frequency was chosen to compensate for the Doppler

effects of all motions of the telescope with respect to

the recombination line emitter and so that the velocities

of the spectra would be with respect to the local standard

of rest (LSR). Since it was expected that the recombina-

tion line emission regions and galactic H 1 regions are

kinematically related, the 21 cm emission velocity lon-

gitude maps of Westerhout (1969) were consulted to deter-

mine the velocity offset needed to position the antici-

pated H157a and C157r spectral features in the 2.5 MHz

spectrometer passband. The velocity offset for the pul-

sar was +50 km/s. All other program regions involved an

off set of -10 km/s. The synthesizer frequencies were cal-

culated by the NRAO computer program DOPSET before any ob-

servations were made.

The 150 MHz IF from each receiver was monitored by

a continuum reciever. The remaining signal was fed to

the NRAO Model II autocorrelation spectrometer. The spec-

trometer creates 413 frequency channels which were divided

into three sections. One section of 192 channels was con-

nected to each of the front-end receivers. The remaining 29-

channel section of the spectrometer was connected to one

of the recievers to serve as a monitor. It was not used

for the acquisition of the data.

Each section of the spectrometer had a passband

of 2.5 MHz. This allowed a channel separation of 13.0

kHz. Each channel has a sin v/v profile with a half-pow-

er width of 15.8 kHz. At a frequency of 1.6832 Ghz, the

velocity resolution of each channel was 2.82 km/s.

The operation of the spectrometer system is de-

scribed by Shalloway et al. (1968) and the Total Power ob-

serving scheme is described in Chapter III. The autocorrelator

periodically transfers the autocorrelation function array

into the on-line computer which performs the Fourier trans-

form and loads the resulting spectrum array onto a seven-

track magnetic tape. This array, along with source coor-

dinates and control information, constitutes a record. The

control information contains a number designation of the

spectrum and whether it is a spectrum of a program (ON)

source or a reference (OFF) source, The identity of its

OFF spectrum is also specified.

Each source is observed for a length of time (tF

or tN as used in Chapter III) during which a sequence of

records is placed on the tape. This sequence of records

is treated as a unit during off-line analysis and is

called a "scan." At the beginning of the off-line pro-

cessing, the records of each scan are averaged together

and the spectrum of each scan is loaded on another mag-

netic tape. Then, for each ON spectrum, its OFF spectrum

is located and the antenna temperature spectrum is com-

puted from

T'(i) = (S(i) R(i)) T

R ( i ) sys

where S(i) and R(i) are the source (ON) and reference (OFF)

spectra, respectively, as a function of the channel, i.

The system temperature, Tys, is computed by the off-

line computer on the basis of a calibration noise source

whose power is periodically injected into each receiver.

The accuracy to which the system temperature is determined

depends on how well the calibration temperature is known.

During the observations the values determined during bench

testing of the receiver were entered into the DDP-116 pro-

gram. The calibration signal was injected into each re-

ceiver by a loop "antenna" placed in the waveguide at a

point directly following the coupler which separates the

two polarizations. During the bench testing the horn was

replaced by a termination. In spite of the careful match-

ing of the termination to the system, its characteristics

are somewhat different from those of the horn observing

the sky. Consequently the amount of calibration noise

power entering the system is not exactly the same in the

two situations (Turner,1973).

The calibration temperature discrepancy was cor-

rected by daily calibrations using the radio source Virgo

A whose continue spectrum is well known. Once a more ac-

curate calibration temperature was known, a correction fac-

tor was later applied in the off-line processing. The con-

tinum flux of Virgo A at 1.6832 GHz is 182 Jy which was

determined by plotting the fluxes published by Kellerman

et al.(1969) and interpolating to this frequency. The con-

tinum receiver connected to the IF was used to measure the

relative levels of Virgo A and the cold sky with and with-

out the calibration signal. Table IV-2 shows the results

of the daily calibration by which the system noise temperature,

T was also determined. The first calibration on 22

sys

April gave unexplainable bad results; so, the calibration

was repeated later. The spectra obtained in the meantime

showed no indication of a problem. The values obtained

during this bad calibration were not used to compute the

averages from which the correction factors were computed.

Before the telescope was pointed toward any source,

the receiver frontends were tuned so that the spectro-

meter passband was located in a portion of the reciever

passband (25 MHz wide at the 3 db points) which was rea-

sonably flat. The daily observing procedure began by peak-

ing up on Virgo A in order to check the pointing accuracy

of the telescope and obtain a correction factor for the

calibration temperature. No problems with the pointing

were found.

Before observing the first program region, a five-

minute OFF scan followed by a five-minute ON scan of M17

was made to check the operation of the spectrometer. The

OFF regions for this and the program regions were at least

four degrees above or below the galactic plane where it

was certain that the H157a line was not present. The H175a

spectrum of M17, an emission nebula, is very strong.

A five-minute scan gave adequate indication that the spec-

trometer was operating properly. The daily M17 spectra

were compared to look for degradation of the system. None

was noticed.

Table IV-2

Calibration and System Noise Temperatures

Date T T T T

Date ys B cal A sys B cal B

April

18 62.2 3.12 58.5 3.17

19 66.4 3.45 59.4 3.98

20 64.5 3.35 60.5 3.55

21 61.9 3.02 59.6 3.75

(22a 176.9 8.57 187.0 7.83)

22 60.4 3.21 58.0 3.21

23 84.5 4.40 84.4 3.59

Average 66.6 3.42 63.5 3.54

"Bench"values 3.23 3.63

Correction factor: 0.975 0.972

aThese are the results of the bad calibration and

are shown for reference only. They were not used to com-

pute the averages.

Most of the observing program for 3C391, its COMP,

3C396, and its COMP was performed with the Total Power

scheme using a sixty-minute cycle; a twenty-minute OFF

scan was followed by a ten-minute scan on each of the

four program regions in succession. The ratio of OFF-

time to ON-time was determined on the basis of Equation

111-5. To make the most efficient use of observing time,

it is desirable to make the scans as long as possible so

that the total time moving between regions is minimized.

The maximum length of the cycle was limited by the length of

time during which the receiver passband remained stable.

During the observations it was found that the quotient spec-

tra of the last few records of the sixty-minute cycle be-

gan to show a curvature of the baseline. This means that

the receiver passband had changed enough since the OFF scan

had been recorded that this OFF spectrum no longer repre-

sented the current passband well enough to produce a flat

continuum on which the spectral features are enhanced. The

curvature only affected the last few 3C396 COMP spectra;

so, the cycle was maintained at sixty minutes.

The observations of these four regions continued

until the RMS noise, expected after combining all scans of

each region, was about O.0020K. Most of the last two days

were used to observe the pulsar and 3C295. Each was ob-

served individually, usually with a fifty-minute cycle

consisting of a twenty-five minute OFF followed by a twen-

ty-five minute ON. The 3C295 spectra showed no spectral

features or abnormal baselines. Therefore, these data will

not be discussed further in this study.

After each day's observations had been completed,

the seven-track "telescope" tape was taken to Charlottes-

ville, Virginia where an IBM 360/55 computer was used to

average the records of each scan and load the resulting

spectra on a nine-track "user" tape for further analysis.

At the same time a summary of records and the quotient spec-

trum of each scan was printed. These tasks were performed

by the first phase of the TPOWER data reduction program

(Cram,1973). The printout was sent to the observatory the

next day. With these "quick-look" summaries, it was possi-

ble to determine which scans had problems which would make

them unusable. On the basis of these reports some scans

were deleted and reobserved on the last day of the program.

After all observations had been made, the final

spectra were produced using the second phase of TPOWER, a

system of routines controlled by a command language. It

was during this process that the correction factor, de-

termined from the daily noise tube calibration, was applied

to adjust the antenna temperature scale. The scans were

summed and weighted by the effective integrating time of

each scan. Then the spectra of the two receivers were av-

eraged together, thereby doubling the integrating time. The

resulting spectra were convolved with a five-channel "box-

car" function and plotted. The plotted spectra were stud-

ied in order to estimate the structure of the baseline.

Curvature of the baseline of each spectrum was removed with

a TPOWER routine which first computes a specified order

Chebyshev polynomial by means of a least squares fit to

a specified portion of the baseline. The routine then

subtracts the polynomial baseline from each spectrum and the

spectra are then plotted in their final form.

A final task was performed in order to facilitate

the analysis of the spectra. Gaussian functions were fitted

to the spectral features. The TPOWER routine begins with

initial estimates of the center channel and half-intensity

width of each feature and, from that, determines a best-

fit center, half-intensity width, and central intensity of

the features.

The Spectra

The basic spectra for the program regions are shown

in Figures IV-2, -3, and -4. Preceding the figures is

Table IV-3 which lists the total effective integration time

(ti) and the order of the Chebyshev polynomial base line

which was removed. The effective integration time was com-

puted with Equation 111-4. It includes the doubling ob-

tained by averaging the spectra of the two receivers.

The RMS noise, listed in the third column, is com-

puted from Equation III-3;

1.528 T

AT = s

P ti

T is the average system temperature of both receivers

sys

and is 65.0 oK. The effective bandwidth, B, is five times

the channel separation of 13.0 kHz because the spectra are

displayed after being convolved with a five channel box-

car function. Table IV-3 also provides a table of contents

which includes a list of the figure numbers for all spectra

including those showing the Gaussian fits to the spectral

features.

The lower portions of Figures IV-5 through 10 show

the results of fitting Gaussin functions through the spec-

tral features. The Gaussian functions of the individual

features are added together and the combined function is

plotted through the original spectrum. The upper portions

of these figures show the plots of the residuals which are

obtained by subtracting the combined Gaussian function from

the spectrum.

The two possible features at -118.8 and 58.2 km/s

on the COMP 3C396 spectrum are only about four times the

RMS. Therefore, it was decided not to present Gaussian

features fit to them.

The center velocity, half-power width, peak antenna

temperature and integrated line power of each feature in the

Table IV-3

Parameters of the Spectra

Basic

SOrder of T Spectra Gaussian Fitting

i RMS Baseline Baseline C Figuresb Compo- Figure

Source [hours] Noise [oK]a Removed [K] (Chapter IV) nents (Chapter IV)

3C391 7.12 .0024 2 10.7 IV 2 A 2 IV 5

3 6

3CO3 6.49 .0026 2 6.1 2 B 2 7

3C391 3 8

3C396 6.66 .0025 4 8.8 3 A 1 9

COMP

3C396 7.08 .0024 2 4.5 3 B

Pulsar

1858+03 7.10 .0024 2 3.7 4 2 10

tion.

After smoothing with a 5-channel boxcar and averaging both receivers.

The designation A refers to the upper portion of the figure, B to the lower por-

spectra are listed in Table IV-4 which follows the spectra.

Also listed for each parameter is its standard error (stan-

dard deviation of the mean) which, except for the standard

error of the integrated line power, is computed by the

Gaussian fitting routine. The standard deviation of the

line power, sp, is computed from the following equation

(Parratt 1961, p. 117) with Equation 11-35:

2 2

P 2 2 (P 2 1/2

sp = [( ) + ( )2s ]

P 5T T 9Av Av

2 2 2 2 1/2

S1.0645 [(A\)2 s2 + T s A1]

Here, sT is the standard error of the peak line temperature

and s A is the standard error of the line half-power width.

The discussion of the spectra will be a part of the

analysis of the emitting region. Chapter V will present

the analysis of the spectrum from the directions of 3C391

and its COMP beam. The model derived there will be applied

to the analysis of the medium in the direction of 3C396,

its COMP beam, and the pulsar 1858+03.

Figure IV-2. H157a Spectra from the Direction

of 3C391 and Its COMP Beam. The

velocity resolution is 11.6 km/s.

0.06 3C391

0.04

0.02

0.00

T'

K

0.06 COMP

3C391

0.04

-170 -130 -90 -50 -10

30 70 110 150

v km/s

____

Figure IV-3. H157a Spectra from the Direction of

3C396 and Its COMP Beam. The veloc-

ity resolution is 11.6 km/s.

-170 -130 -90 -50 -10 30 70 110 150

v km/s

0.022

0.014

0.006

0.002

T'

K

0.022

0.014

0006

-0.002

Figure IV-4. H157a Spectrum from the Direction of the Pulsar 1858+03.

The velocity resolution is 11.6 km/s.

1858+03

-120 -80 -40

0 40 80 120 160 200

v km/s

0.011

T'

0.007

oK

0.003

-0.001