..

O

M

01“, u . .

can... J\be:.~. . it“; I

" JiliaTOUO‘dsuuo-Jzfl: u v,

. ‘ h-O 1‘94“ . ‘ L.“ x 3 .'“

“""" .10" 3.; '”

V‘ -

‘4‘. W0

‘2’...“ E.“ '1'. OW . J
5 b. - .7- _w¢§.‘:: alo-

  
 
 
 
  
    
  
    

  
   
 

 

   

.- "3 7
u;f..«~KM-JJ “figm'agdxa'fin
gr I V «a a. - ..- .7 I
L! ”“3“ J. \Q‘.’ .9373": 0‘” 0:7.
“1}?b-.., *- w“ -‘é‘x.
-.' |~._pu ‘1‘.
5 w... ..,. I 3 .
17.97.
I .

    
    

       

 

 
   
   
  

00-,“ ‘ ‘ '
7-.»
>41 . tho'r‘ .
.D‘D".H. ”"NB'
. ..",|.,u & «y
'- r § .1 ‘ 4‘
M’ I a
.4 9 ‘g’t a-O 9"
' "' -.J:.r .
O‘- .

  
 

   
   
  
   
 
   
    
  

    
 
 

  
 
 
   

I
.a- .
' 53:5
"a ---.*1- ,p

*‘x‘fix'kfgéirx “£38 1 '3"!
‘ ’ 'Wv .,

° 3% riéfiw'ks,

“m“ ‘ . on: ‘ I

 
  
  
 

 
   
 

 
   
 
   
 

      

. ' 0.. .~
”figs. ..
.1.

.‘..7 v.-
4 l”-

H.-

. .d. ~

0 \I. 4,. u . ‘n.‘ ~‘ .‘

' ‘ “EIQpJ-.?.-' .r
‘“?“.' I

Lt I
~11. .

  

‘ “-5 .
.v .f-nuon-
”01..

 
  
   
    
    
  
    
 
  
   
   

  
  

  

w;ru¥gc’nwc _
Ola-1‘1”“ "£0" . .
o ' CPA -IO"

.aS'é‘fi‘Wi

  
  

    

‘a- 9-0 30“
6‘0! Q
on

 

 
    
 

 
   

1.
all-

h' v
. I;

 
 
 

O

I

‘l' "n
.34. — in
“.fjtman ya-
“ : .. A- L.

'u‘.«'4~‘

«yup a
N.

l

   

  
  
  
   

‘

' I‘40f.

- - Wfl-‘ ‘1
"2““ he". ' ~‘

.'
4
‘9'
K

 

      

   
   
  
  

 
  

“4 {on
II‘ I .0

r
0“

    
  

1‘1 .

        

. . '
' . Q_
' u .
I
I - . t u _ -
. o | . h
. . -. U ‘ '. : . l”
g . .5 . ‘ . . I- f
I . 1. -Q '. ' _ ; v‘ .. .. . - '
~ ‘ , ' . .l ‘ : ’ . q‘
l-,‘ o ¢ , I O.. . I , . ‘. ta...
. -- I I , ' O ‘ ' .
t .. 0 . ' g .4! . l .o .I. _
. . . ' v .
- - _‘ o '. .-" 1 gr
9 n 'v . o ' ' . - ‘-
. . '.
a V . V . . ‘ o‘.‘
o n . - - .0 l d .
‘ — n . - ‘ .
. ' ' "N w , ‘
.. g ‘ I .~ . _ I. "’4
. .1 I . . ‘ . ‘ '
t . . 3- L - o o " _ ‘.i
C ' . "
o 0 ' ‘ ' I I
V. . ’ ' _ . -S _ .v .,‘.'.v
o o t . v‘_ _ 'u_.. . ' “'r‘. .
. ,no 0 ‘ . , ' O p.‘! 'J".‘3'--_.‘
' ' h— 'I I
\

 

 

LIBRARY 1
Michigan State

University i

J

 

 

—__—

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before due due.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

F 1i? #1

* MSU Is An Affirmative ActiorVEquel Opportunity Institution
7 Wfllfi-pd

 

 

A COMPARATIVE STUDY OF THE IDENTIFICATION OF

MIDRANGE PETROLEUM PRODUCTS BY

GAS CHROMATOGRAPHY AND 3-D FLUORESCENCE SPECTROSCOPY

By

Cheng Nanzheng

ATHESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE

School of Criminal Justice

1991

ABSTRACT

A COMPARATIVE STUDY OF THE IDENTIFICATION OF
MIDRANGE PETROLEUM PRODUCTS BY
GAS CHROMATOGRAPHY AND 3-D FLUORESCENCE SPECTROSCOPY
By

Cheng Nanzheng

Twenty-two mid-range petroleum-based products, including charcoal lighters,
paint thinners, and synthetic solvents, were studied by capillary GC and 3-D
fluorescence. It was found that 3-D fluorescence is far better at discriminating
among similar products than was GC, which could only put the products into
broad classes. When partially or totally evaporated, the 3-D fluorescence plots
were altered enough that they could not be matched with those of neat
samples. A similar situation exists when they are burned. Also, when burned
even under controlled conditions, these compounds did not yield consistent

fluorescence patterns.

A blind test on neat samples showed that 3-D fluorescence can be a reliable
technique for determining whether or not two neat samples were of the same

brand.

ACKNOWLEDGMENTS

With greatest respect and gratitude, the author would like to acknowledge Dr.
Jay A. Siegel, the professor at the School of Criminal Justice, Michigan State
University, who was cordially and consistently directing the author to conduct
all steps of the study - from the choice of topic to the experiment design, from
the grant obtaining to the formation of this paper. Without his considerate
guidance and warmhearted help, the author could not move ahead, or could

stop at any stage of the project.

With great gratitude, the author would like to acknowledge Dr. Frank
Horvath, Dr. John Hudzik, at the School of Criminal Justice, Michigan State
University, and Lt. Roger J. Bolhouse at the Crime Laboratory of Michigan
State Police. All of them offered me a lot of very valuable suggestions in the

formation of this thesis.

Furthermore, the author would like to thank to the Midwestern Association of
Forensic Scientists and the American Society of Crime Lab Directors for their
financial support in this project. The author would also like to thank Mr. Rick
Tontarski and Ms. Patricia Pannuto of the Bureau of Alcohol, Tobacco, and

Firearms for furnishing some of the samples used in this study.

CONTENTS

PRECLUDE

PART I. INTRODUCTION
1. ARSON CRIMES IN GENERAL
2. BASIC TECHNIQUES OF ARSON INVESTIGATION
2.1. ARSON SCENE INVESTIGATION
2.1.1. Recognition
2.1.2. Collection
2.1.3. Packaging and Delivering
2.2. LABORATORY METHODS IN GENERAL
2.2.1. EXTRACTION
2.2.2. ANALYSIS
2.2.3. CHEMICAL ANALYSIS
2.2.4. INSTRUMENTAL ANALYSIS
3. GAS CHROMATOGRAPHY
4. THREE DIMENSIONAL FLUOROMETRY

PART II.RESEARCH DESIGN

iv

10

12

13

16

17

18

19

20

21

21

25

32

46

PART III.

PART IV.

PART V.

MATERIALS AND INSTRUMENTATION CONDITIONS
SAMPLES

FLUORESCENCE SPECTROPHOTOMETER

AND ITS CONDITIONS

GAS CHROMATOGRAPH AND ITS CONDITIONS

PROCEDURES AND RESULTS
STAGE I. NEAT SAMPLE GROUPING
STAGE II. CONCENTRATION STUDY

49

49

50
54

56
56

80

STAGE HI. BLIND TEST TO COMPARE TWO SAMPLES 88

STAGE IV. EVAPORATED SAMPLES
STAGE V. BURNED SAMPLES

SUMMARY

REFERENCE

94

104

115

117

PRELUDE

Arson investigation has been one of the most challenging subjects in the field
of forensic sciences. The detection and analysis of accelerants are always
considered one of the central themes in arson investigation, and has been the
subject of many books and articles over the past thirty years or so. Numerous
kinds of materials can be used as an arson accelerant. The most common ones
are petroleum distillates or products. A sub-class are the so-called midrange
petroleum products such as charcoal lighters and paint thinners. Because they
are commonly found around the home, they have the potential to be used as
an accelerant in arson fires. However, it is somewhat surprising that the
midrange petroleum products have not been the subject of much research into
classification and methods of detection and analysis. Therefore, the author
believes that a careful study of midrange petroleum products will contribute
to arson investigation in general as well as to petroleum-based accelerant
analyses in particular. This is why Dr. Jay A. Siegel suggested the author

choose this topic as his Master’s thesis.

In Part I of the thesis, the author intends to give a broad but brief picture of
arson crime in its legalistic and technical aspects; then change the focus from
arson investigation to general laboratory methods of accelerant analysis, and

finally to highlight the gas chromatography (GC) and three dimensional

2

fluorescence spectroscopy techniques which were used in this study.

In Part II, III VI and V, the study involving twenty-two midrange petroleum-
based products, including charcoal lighters, paint thinners, and synthetic
solvents, is described. All the samples were measured under the conditions of
(1) neat state, (2) 50% evaporated state, (3) 100% evaporated state, and (4)
totally burned state. In addition, a concentration study and a blind test of neat
sample were also conducted. It was found that 3-D fluorescence is superior at
discriminating among similar products than was GC for neat samples. When
partially or totally evaporated, the 3-D fluorescence plots were altered enough
that they could not be matched with those of neat samples. A similar situation
exists when they are burned. Also, when burned even under controlled

conditions, these compounds did not yield consistent fluorescence patterns.

PART I. INTRODUCTION.

1. ARSON CRIJVIES IN GENERAL.

N 0 person can argue with the fact that arson is one of the major property
crimes in the United States as well as in most other countries in the world.
Its seriousness could be viewed from difi‘erent perspectives. First, in terms of
crime rate, although the incidence of arson is not as high as the crime of
robbery, auto theft, or breaking and entering, it is comparable in number to
the homicides committed each year [1]. Second, in terms of arson occurrence,
in 1975, the report from the National Research Committee of Fire Research
estimated that perhaps as high as 40% of all fires are caused by arson [2].
Third, in terms of amount of money lost by the victims, the above report has
established the annual loss from wrongful setting of fires at between 5 and 6
billion dollars. Fourth, in terms of death rate, it was estimated more than

4,000 deaths from arson occur each year in the United States [3].

In the United States, historically, common law conceptions of arson referred
to the "malicious burning of the dwelling of another". However, modern
statutes have altered the notion of arson in variety of ways. For example,
Massachusetts has expanded the definition by passing legislation defining

arson as "the willful and malicious setting fire to, or burning of, any building

4
or contents thereof even if they were burned by the owner" [4]. According to

the Uniform Crime Reports (UCR), which publishes nnually by FBI, arson was
assigned as item 8 of the Part I ofl‘enses ("crime known to the police and
arrests"), and was defined as "any willful or malicious burning or attempt to
burn, with or without intent to defraud, a dwelling house, public building,
motor vehicle or aircraft, personal property of another". Today, like most other
legal concepts, the definitions of arson are not the same in different states.
However, in comparison with the common law, the expansion of modern
statutes about arson at least could be summarized in three aspects. (1) While
arson originally carried the ideas of fire and burning, many jurisdictions now
include the use of explosives. (2) Contemporary statutes include not only
dwellings, but also other types of buildings, as well as other property like
motor vehicles and aircrafi: as the subject of the arson crime. (3) The property
of the arsonist is also included if there is an attempt to defraud an insurer or

if the building is occupied [5, 6].

Though the legal definitions of arson vary historically and geographically, it
was and is a felony in all jurisdictions. Typically, arson is divided into at least
two degrees and sometimes three [7]. In general, if the premises set afire are
occupied, the charge will be "first-degree arson". If they are unoccupied, the
case will be one of "second-degree arson". A person is guilty of arson in the

third degree if the premises burned are his or her own, if they are unoccupied,

5
and if the purpose is to defraud the insurer.

Arson is a complex crime and often presents difficulties during investigation
and prosecution. Take Connecticut State as an example, in the single year of
1980, there were 6820 incendiary and suspicious fires, but only 500 arrests
(about 7.3%) were made, and only 72 people (only about 1.1%) were convicted

of arson [8].

The complexity of arson cases can be viewed fi'om two major perspectives:
legalistic and technical. Viewed from the legalistic perspective, the principal
problem is the element of criminal agency, or intent. Studies have shown that
the reasons for arson are both numerous and obscure, and sometimes hard to
detect. For instance, there are "revenge firesetters", whose crimes result from
anger, hatred, or jealousy in personal relationships; there are "excitement
firesetters", who simply enjoy watching fires and the operations of fire
equipment; there are "insurance-claim firesetters", who incinerate business
property for its insurance value; there are "vandalism firesetters", who set
building ablaze as part of adolescent peer-group activities; there are "criminal
vindication firesetters", who use arson for hiding the evidence of other crimes.
There also might be "professional torches", who incinerate buildings for a fee,
the "firebomers" of activist and political liberation organizations, the large

number of skid row vagrants who seem to be over represented among those

6

arrested for arson, and the many other types for whom intent and motivation

are not altogether clear.

From the technical perspective, a fire can have many accidental causes,
including faulty wiring, overheated electric motors, improperly cleaned and
regulated heating systems, cigarette smoking, etc.. The ultimate task of fire
investigation is to establish the willful or malicious setting of a fire. Therefore,
final determination as to the cause of a fire must take numerous factors into
consideration and requires an extensive on-site investigation as well as
laboratory analysis to characterize the physical evidence which could link the
cause(s) of a fire and link the person’s intention of setting a fire. This task
presents a major challenge with many troublesome circumstances to
investigate and analyze. First, arson is likely to be conducted premeditatedly
with a thorough plan and is committed in secret; little direct evidences or eye
witnesses are normally available. Second, it is ordinarily done at the
convenience of a perpetrator who has left the crime scene long before any
oficial investigation is launched. Third, all kinds of potential physical
evidence are rendered more difficult to obtain than most other kinds of crimes
because of the extensive destruction by the fire. Fourth, the top priorities on
a scene of a fire are the fire-fighting and life/property rescuing operations
rather than collecting potential crime evidence in a most careful and time-

wnsuming manner. Notwithstanding, the first task definitely and sometimes

7
devastatingly influences, disrupts, or damages the success of the second task

in an irreversible way.

All the above factors indeed attribute to the difficulties of arson investigation,
but they do not mean that it is impossible to get physical evidence for
prosecution of arson cases. As a matter of fact, modern technologies have
demonstrated that by means of well-organized, scientific procedures and a
thorough search at the fire scene on the one hand, and by means of well-chosen
analyses in crime laboratories on the other hand, various types of physical

evidence can still be found and can yield valuable results.

The types of physical evidence at an arson scene depend on the method used
by the arsonist to set the fire, and can be placed in two broad categories: (a)
physical evidence which may be associated with the suspect but not with the
incendiary itself, and (b) residual components of the incendiary. The first
category includes fingerprints, shoe impressions, toolmarks, glass fragments,
blood, tissues, hairs, fibers, and many kinds of trace materials. Though in a
particular case, one of these types of evidence may be collected and be crucial
in investigating or in prosecuting procedures, because of the destructive nature
of fire itself and the chaotic condition of fire scenes, most evidence in this
category is likely destroyed or lost, hence infrequently acquired by scene

investigators. The second category -- residual components of an incendiary, in

8
its broad meaning, includes ignition devices, matches, cigarette lighters,

papers, pieces of wood, and a variety of accelerants which can assure the rapid
start, the enhanced intensity, and the extent of a fire. Among them,
experience has shown that accelerants are involved in most arson cases and
are frequently found at arson scenes. Therefore, in suspected arson cases, the
forensic scientist’s function is primarily concentrated on the detecting,
demonstrating and analyzing accelerant. The expert testimony of identification
of accelerant is extremely important to the success of the prosecution because
the discovery of an accelerant establishes "intent" of the arsonist -- a

requirement under most statutes as we discussed above [9].

According to its physical state, accelerant can be classified into three groups:
(1) gaseous state, (2) solid, and (3) liquid. Gaseous accelerant consist of coal
gas, natural gas, propane, and so forth. Solid accelerant include fireworks,
explosives, and so on. In arson investigation, it is well acknowledged that the
most commonly encountered accelerant are liquids. They include gasoline, fuel
oil, kerosene, charcoal lighter, paint thinner, dry cleaning fluids, wax solvents,
methanol, alcohol, turpentine, lacquer thinner, industrial solvents, and
numerous types of flammable chemical reagents, or any proprietary products
that utilize any of the above as carrier or solvent. Among them, petroleum
distillates or petroleum-based products are estimated to occupy more than 95

% of the total, because they are easily available and are highly flammable [10].

9
In order to correctly analyze a suspicious liquid submitted to a crime

laboratory, it is helpful to have some brief knowledge of the nature and
classification of most frequently confronted acceleran -- petroleum-based
products, or petroleum distillates. They are mixtures of many different
hydrocarbons, especially different types of polycyclic aromatic hydrocarbons
(PAHs). However, the exact components and the composition of these
components of a particular product are often not specified or have not been
fully studied. For the purpose of our discussion, it is convenient to categorize
them into two classes: straight-run products and mixtures of specially

processed ones. The following shows the common accelerantes in each class:

Class I. Straight-run products:
1. Light petroleum distillate; for example, naphtha, lighter
fluid, cleaning fluid.
2. Medium petroleum distillate; for example paint thinner,
charcoal lighter.

3. Heavy petroleum distillate; for example kerosene, diesel
fuel, fuel oil.

Class II. Specially processed petroleum products:
1. Separate aromatic; for example, benzene, toluene, xylene.

2. Blended aromatic and aliphatic; for example, gasoline.

10
3. Blended or modified products; for example, lubricants and

special solvents.

The main points above can be summarized as: (1) Arson involves two major
aspects: legal and technical. (2) Physical evidence is the key part of the
technical aspect. (3) Among various types of physical evidence, residual
materials of incendiary are mostly found. (4) Among various residual
materials of incendiary, suspected accelerant are mostly found. (5) Among
various suspected accelerant, suspected liquid accelerant are mostly found. (6)
Among various liquid accelerant, petroleum-based products are most often
found. Therefore, the next section will focus on basic techniques of detection

of liquid petroleum-based products in arson investigation.

2. BASIC TECHNIQUES OF ARSON INVESTIGATION.

In the situation of most other criminal investigations, when the detective work
starts, though there are a lot of unknown things, at least one thing is clear --
it is a crime. But in most fire cases, when the scene investigator starts his
work, the cause of the fire is still likely unknown. Therefore, three questions
should be answered through the fire investigation and prosecution, if possible.
The first is "HOW", i.e., how the fire started? In other words, was the fire an

11
accident or was it set intentionally? The second is "WHO", i.e., if the fire was

set intentionally, or it was an arson, who started the fire? The last question
is "WHY", i.e., if the person who set the fire had been found, why he or she did
it? The "HOW" question is mainly answered by scene investigators. The
"WHO" question is mainly the result of a joint efforts of scene and laboratory
personnel. The "WHY" question is mainly solved through the court inquiry
process. This last question, though which is critical in the trial and sentencing
process, is out of the scope of our study and will not be discussed in the rest

of this paper.

Usually, in the stage of answering "HOW" and "WHO" questions, a suspicious
liquid (Sample A) is detected and identified at or around the fire scene. Latter
on, in addition to Sample A, a control liquid (Sample B) which is possessed by
the suspect(s) is frequently submitted with the request that the laboratory
determine if it came from the same source as that identified in the scene. In
a case like this, there exist two tests: (1) a necessary two-step detection of
Sample A and Sample B, and (2) a potential two-step physical Link; between the

fire scene and an individual.

For the detection task, the first step is to find Sample A at or around the fire
scene, and the second step is to find Sample B. Either of these steps could

only be performed by experienced professional through a careful and thorough

12

search, and then be well preserved and delivered.

For the link task, the first step is establishing the link between the liquid and
an individual. This is usually the responsibility of the submitting agency and
is accomplished by direct possession, fingerprint comparison, and so forth. The
second link, that the link between the Liquid A and Liquid B, is more difficult
to achieve, and is primarily the responsibility of crime laboratory. The
variation of individual products within the pertinent general product
classification must be determined before the attempt to establish this second

link [11].

2.1. ARSON SCENE INVESTIGATION

As with other types of criminal investigations, the key to successful
examination of arson evidence begins with the scene investigator. The final
determination of whether a given fire is arson usually depends upon the skill
and experience of the fire scene investigator; however, the forensic science
laboratory can assume an important backup role by determining the
flammability and the identity of an accelerant which is recovered from the fire.

Arson-scene investigation, which is often long and tedious, and it generally

13

includes:
1. Determination of the point of origin of the fire.
2. Examination of electric and mechanical equipments to evaluate
the possibility of the fire having been caused by their failures.
3. Examination of the burning patterns.
4. Recognition and collection of the relevant physical evidence.
5. Analysis and reconstruction of the scene.

6. Determination of the cause of the fire.

In all steps of the above procedure, a question about the possibility of arson
should always be kept in a scene investigator’s mind. Particularly, the above
steps 3 and 4 are key processes if the fire is suspected as an arson. When
suspected liquid(s) are involved, the following three approaches are needed to

be conducted properly in order to get proper samples for laboratory analysis.

2.1.1. Recognition.

Though as it has been emphasized above that arson investigation provides an
exceedingly difficult circumstance to recognize and preserve evidence,
fortunately, practice shows that only under the most ideal of conditions will

combustible liquid(s) be entirely consumed during a fire. For example, when

14
a flammable liquid is poured over a large area, it is highly likely that a portion

of it could seep into porous surfaces nearby, such as cracks in the floor, plaster,
upholstery, rags, wall boards, or carpet, where enough of it remains unchanged
so that it can be collected by scene investigators. On the one hand, the liquid
seeping naturally tends to go downwards by the force of gravity; on the other
hand, the center of combustion is in gaseous state, and naturally tends to move
upwards because of the effect of hot air causing lower pressure, and tends to
sweep over towards a more ventilated direction very rapidly, consequently
leaving the original point of fire below or behind. The results of these two
opposite movements provide a considerable chance for scene investigator to
discover liquid accelerant residues even in the situation where a fire was very
serious. In addition, when a fire has been extinguished with water, the rate
of evaporation of volatile fluids may have been slowed down because water
cools and covers materials through which the combustible liquid may have
soaked. Favorably, the presence of water does not interfere with later
laboratory methods utilized to detect and characterize flammable liquid

residues unless there are huge amounts of water.

Arson with a flammable liquid usually produces a characteristic structural fire
pattern which will give the investigator some hints in identifying promising
areas, such as the origin of the fire, for evidence collection. Ifthere are some

porous or absorbent materials near the origin (such as paper, carpeting, fabric,

15
book, etc.), these may have retained a substantial amount of volatile accelerant

and should be collected even no visible liquid could be found by the naked eye.
However, when solid state materials such as paper, pieces of wood, or trash are
solely used to start a fire; or when gas phase materials such as natural gas or
propane are used as an arson tool, there will usually be little or no residual
accelerant for laboratory analysis and the burden of proving incendiary must

be borne by the fire investigator on the scene.

In recent years, many scene investigators have begun to use flammable vapor
detectors or chemical tests to help in locating trace amounts of accelerant at
the area of origin. There are several types of portable flammable vapor
detectors available commercially. In general, field tests for accelerant residues
can be placed into five categories, according to their principles of detection:

(1) chemical color tests,

(2) catalytic combustion detectors,

(3) flame ionization detectors,

(4) portable gas chromatographs, and

(5) portable infrared spectrophotometers.

The evaluations of these means have not achieved the point to rank them.
Nonetheless, it is safe to say that all these techniques are helpful to a certain

degree. However, up to now, visual inspection by experienced scene

16
investigators is still the most productive and reliable means to detect the

location of a flammable liquid.

2.1.2. Collection.

Even if any petroleum-based product residues remain after the fire is
extinguished, the most valuable and substantial part of them for analysis will
evaporate within a matter of hours and certainly within a couple of days.
Furthermore, mopping up and the overhauling operations by the fire fighters
will dramatically decrease the chance to conduct a meaningful investigation of
the fire scene. Therefore, the evidence recognition should be started as soon
as the investigator has arrived at the fire scene, and the evidence collection
should be launched as soon as the suspected materials have been found
provided his work will not influence the major fire-fighting and life-rescuing
operations. Unnecessary moving of burned debris, furniture, and other goods,
as well as excessive walking around at the scene, will alter the scene and make

the investigation more difficult.

In addition to their simple use as accelerant, flammable liquids are used in the
construction of firebombs or "Molotov Cocktails". A Molotov Cocktail consists

of a flammable liquid, usually one of a low flash point such as gasoline or

17
charcoal lighter, in a breakable container and fitted with an ignition system.

When such a device is used, fiagments of the container or portions of the wick
may remain around the fire origin and should be collected for laboratory

examination.

2.1.3. Packaging and Delivering.

Petroleum-based products are highly volatile. Hence, when suspected liquids
or materials which could contain an accelerant are collected, improper
packaging and delivering can lead to total loss of fire residues. Some
precautions should be in mind of scene investigators:
1. There are three crucial periods of time, one is the time between
recovering suspected material(s) and putting them into containers
(i.e. packaging time), another is the time between putting them
into containers and sending them to crime laboratory (i.e.,
delivering time), the third is the time between they arriving to
the laboratory to conducting the analysis (i.e., the storaging time).
All these three periods of time should be as short as possible.
2. If any material which is suspected containing flammable liquid
(often called "flammable liquid carrier") is found, it should be cut
off from other part of the material, and should be transferred into

a tightly sealed, clean, and previously unused glass jar or a metal

18
can, and marked with a label containing information such as the

investigator’s name as well as the date, the department, the
location, the area fi'om which the sample was taken, case number,
and description of the item. Paper bags or envelopes are not
suitable due to their porosity. Zip lock plastic bags and most
kinds of other plastic containers are also unsuitable because
possible chemical reactions can change the nature of petroleum-
based products and ruin the whole analysis.

3. Along with the suspected flammable liquid carrier, it is absolutely
necessary to get "control evidence" by cutting pieces of the same
material fi-om an area which does not contain flammable liquid
residues.

4. If the suspected materials can not be examined immediately in
the crime laboratory, they should be put into refrigerator at the

temperature around 4° C as soon as possible.

2.2. GENERAL LABORATORY METHODS FOR ACCELERANT
ANALYSIS.

If it is believed that the final determination about whether a given fire is arson

depends upon the work of scene investigation, it is also believed that the final

19
determination about whether Sample A could be linked with Sample B depends

upon the work of laboratory analysts. Moreover, the laboratory work is

considered more important in terms of the success of case prosecution, and

more difficult in terms of the techniques available.

2.2.1. EXTRACTION.

Normally, the suspicious sample which collected by scene investigators (Sample
A) is not a pure liquid, but is in a complicated state -- various kinds of porous
materials such as rug, paper, cloth, wood, insulation and so forth usually go
with it. Therefore, before a meaningful analysis could be carried out, it is
necessary to recover a potential accelerant from its matrix in the form of a
liquid or gas. This is the first and most critical step of analysis. A variety of
means have been developed for this purpose. At least five of them have been
implemented in conventional forensic practice: (1) distillation, (2) solvent
extraction, (3) water flotation, (4) vapor concentration, and (5) head space. For
the small amounts of the residual flammable liquids present in most arson
evidence, simple distillation, solvent extraction, vapor concentration or water
floatation techniques frequently fail to retrieve sufficient amounts of an

accelerant to allow identification. Today, in forensic laboratories, the most

20

widely used concentration method with arson evidence is the head space
technique [12]. Basically, this technique utilize a hermetically sealed system
with two parts. The suspicious accelerant with its matrix is put into the first
part, it is usually a glass or metal container. In order to minimize the chance
of sample loss, the original container submitted by scene investigators can be
directly used. Then it is sealed and heated moderately. The possible
accelerant will be evaporated, introduced to the second part of the system, and
reduced into its liquid state in lower temperature. For some analytical
techniques illustrated later (for example, MS or CO), the first part of the

system can be directly connected with the sample entrance of the equipment.

2.2.2. ANALYSIS.

After the suspicious sample has been extracted from its carrier, the next and
the key stage -- sample analysis -- can be conducted. Traditionally, the
analysis of petroleum-based products in the forensic laboratory can be
classified into two basic branches -- physical analysis and chemical analysis.
Physical analysis includes the determination of flash point, distillation range,
reflective index, specific gravity, and so on. These physical parameter
measurements have some limitations. First, considerable amounts of sample

are required which is usually hard to abtain in real arson cases. Second,

21
during most of these physical measurements, the sample will be consumed or

altered, therefore decrease the possibility of successive work. Third, they are
relatively coarse in terms of the discriminatibility among types of petroleum-
based products. Hence, though they had been extensively studied and were
quite popular twenty or even ten years ago, they are getting less and less

employed in routine work of accelerant analyses.

2.2.3. CHEMICAL ANALYSIS.

Chemical analysis of petroleum-based products in arson cases is typically
further divided into two major categories -- chemical reaction analysis (for
example, color test) and instrumental analysis. Customarily, chemical reaction
analysis also needs a relatively large amount of sample, consumed or altered
sample during the tests, and have even less discriminating power than
physical analysis in general. So, only a few forensic scientists are still

routinely using them.

2.2.4. INSTRUMENTAL ANALYSIS.

Instrumental analysis, along with the rapid development of micro-electronics,

22
precision optics and computer techniques, is continuously changing.

Accordingly, though it demands ample investment of both fiscal and human
resources, is still drawing more and more attention in nearly all aspects of

forensic sciences as well as in accelerant analysis.

A great deal of effort has been devoted to the characterization of petroleum
samples by a variety of instrumental techniques. Petroleum-based products
have very complicated components, and can be investigated from different
aspects of their features. Certain analytical methods inspect certain sets of
these characteristics. Consequently, in theory as well as in practice, no one
technique can completely take the place of another. By scrutinizing research
literature of liquid accelerant analysis, it is found that the following
instrumental techniques are frequently employed and can provide
complementary and corroborating information:

(1) Gas chromatography (GC).

(2) Liquid chromatography (LC).

(3) Thin layer chromatography (TLC).

(4) Gas-liquid chromatography (GLC).

(5) Mass spectrometry (MS).

(6) Gas chromatography coupled with mass spectrometer (GC-MS).

(7) Infrared spectroscopy (IR).

(8) Ultraviolet spectroscopy (UV).

23
(9) Fluorescence spectroscopy (FS).

(10) Gas chromatography coupled with infrared spectrometer (GC-IR).
(11) Atomic absorption spectrophotometry (AA).
(12) X—ray diffraction spectroscopy (XRD).

(13) Nuclear magnetic resonance spectroscopy (NMR).

Among the above methods, GC, LC, GLC, and MS are usually utilized for
separating and distributing different types of hydrocarbons or certain frictions

of hydrocarbons (e.g., by MS) in the mixtural sample.

For improved characterization of a suspected flammable liquid both IR and UV
spectroscopy have been explored. A few investigators have claimed success
with these techniques but they are not widely used. UV spectroscopy offers
little discrimination between types of petroleum distillates, as for example,
between kerosene and fuel oil, because both are primarily mixtures of
saturated aliphatic hydrocarbons [13]. IR spectroscopy could offer more
information, primarily in the detection of aromatics [14], which are found in
relatively high concentrations in gasoline as compared to un-cracked products
such as kerosene. No doubt, both IR and MS have high sensitivity (whcih is
defined as the ability to detect minimum amount of sample) and high

selectivity (which is defined as the ablility to distinguish different while

24
similar samples or components of a sample in terms of certain features which

the instrument is targeting), but the production of meaningful spectra requires
pure samples. In most arson cases, pure samples are seldom available even
after using an extracting procedure. The introduction of GC-IR or GC-MS in
recent times, has made it possible to utilize both IR and MS on casework
samples. The mixture is first separated by GC, and the components are then

individually analyzed by IR spectrophotometry or by MS all in one process.

TLC has been very useful to detect dyes which are usually added in many
petroleum products by manufacturers as identification marks of a particular
brand or a particular batch of a brand. TLC techniques are uniquely
successful to separate, segregate, arrange, compare and identify components
of the mixtures of dyes. TLC’s are much less expensive and easy-to-
manipulate than any other techniques listed above, and it is both sensitive and
discriminative to dyes. However, it is not a method which directly probes the
components of petroleum-based products, and not all petroleum-based products
contain dyes, especially in foreign countries such as in Austria and many Asian

countries [15].

Atomic absorption spectrophotometry (AA) and X-ray diffraction spectroscopy
(XRD) are valuable in quantitatively detecting the inorganic elements such as

lead and bromine in petroleum-based samples, especially in gasolines.

25
Some of the above techniques have been refined to the point where they can

be applied to the typing of unknown petroleum samples. Analytical data
employed for petroleum-based product classification include gas
chromatography [16-21], infrared spectra [22-27], the combination of the two
techniques [28], trace metal concentrations [29,30], luminescence spectrometry
[31], field ionization mass spectrometry [32], high pressure liquid
chromatography [33], thin-layer chromatography [34], and other methods [35-
39]. Several of the analytical methods have been standardized and published
as Standard Methods by the American Society for Testing and Materials,
including gas chromatography, elemental analysis, fluorescence analysis, and

infrared spectroscopy.

Though different instrumental analytical techniques can provide
complementary information, GC is regarded as the most powerful one in

petroleum-based product analysis by most forensic scientists.

3. GAS CHROMATOGRAPHY

Though gas chromatography has been extensively used as a major chemical
separation method for years, its mechanism has not been thoroughly

understood. Nevertheless, briefly speaking, it involves the segregation of

26
substances based on their relative affinity for two phases, one is stationary, the

other is mobile. Substances or components of a substance which have higher
affinities for the mobile phase are moved, or carried along with it faster, and
are thus separated from those with higher affinities for the stationary phase.

Basically, a GC consists of three portions: separation, detection and recording.

Early GC workers in 19608 mainly relied on packed columns for separation,
thermal conductivity detectors, and linear recorders. Their results proved that
the presence of hydrocarbons could be detected in fire residue [40]. Recent
instrumental development has enormously elevated GC’s capability in
accelerant analysis. For separation portion, capillary columns up to several
hundred feet are routinely used and offer a resolution of hundreds of peaks of
petroleum-based products. Computerized temperature programming increases
the separation range, sharpens peaks and shortens analysis time, particularly
with higher boiling hydrocarbon fi'actions which are often the important part
of fire residues. For the detection portion, a variety of detectors have been
developed, and a modern GC system is usually equipped with several different
types of detectors, such as flame ionization (FID), electron capture (ECD), as
well as thermal conductivity (TCD), to optimize different applications. Among
them, flame ionization detectors are ideal for arson evidence because they have
very high sensitivity to hydrocarbons and hydrocarbon-related products such

as alcohols. For the recording portion, log recording techniques have been

27
developed and applied in petroleum products [41]. It transfers the ordinary

linear recording mode of Y-axis (usually represent the intensities of GC peaks)
into logarithmic mode, and therefore, tremendously expands the inner-
dynamical extent of a gas chromatogram by increasing the area of weak peaks
and decreasing the area of strong peaks. In fact, weak peaks usually offer
important and substantial information about different samples among a
category of petroleum-based products, such as different brands of paint
thinners. Based upon these new advancements, later work [42] explored the
recovery of many different accelerants from various types of fire residues,
though in recent years, with the development of automatic-interpreters, log
presentation has lost its popularity. Basic parameters used to distinguish
different groups of fire residues include aliphatic and aromatic hydrocarbon
content, relative concentration of major versus minor components, and even the
presence of additives. G0 has also been used to compare petroleum-based
products within a given group [43-45]. In these comparisons, the relative peak
intensities are used to distinguish products containing the same components
but in different quantities. By this approach, products with only a few
components will not yield as much comparative information as products
containing hundreds of components. Fortunately, in the case of petroleum-
based products analysis, modern GC can easily detect hundreds of peaks,
which makes the comparisons ready to conduct. Two comparison mechanisms

were employed in these exercises [46]. The first is a simple qualitative method

28
-- the overlay method. In this method, two chromatograms, with the relevant

peaks on scale and approximately the same size, are simply superimposed on
a light box. Large to moderate differences between peak ratios in the two
chromatograms are readily visible. Extra care should be taken when
comparing sets of peaks which do not separate completely. A slight increase
in separation can cause a significant decrease in peak height. The overlay
method provides a rapid screening procedure to determine if it is necessary to
proceed to the more precise quantitative comparison. The second method is
peak areas comparison which is provided automatically by many new model
GCs and can be used to perform quantitative comparisons. Both of the

comparison methods were adopted in this study.

In addition to its well-recognized selectivity (the separating ability), high
sensitivity is GC’s another primary advantage. Petroleum-based product vapor
as low as a few parts per million in concentration can be detected and
identified quickly and accurately, far surpassing most other techniques known
to criminalists. The third advantage of GC is its extensiveness. It can cover
the entire range of petroleum distillates - from the very light fi'actions of
petroleum (e.g., ether), to the middle distillates (e.g., naphthenes and
kerosene), to heavier products such as fuel oil and lubricating oils. Each of
these is a complex mixture of saturated and unsaturated aliphatic and

aromatic compounds.

29
Due to its peculiar advantages, GC is unparalleled in determining the

existence and identifying the type of petroleum-based products in arson cases.
In 1967, Bruce V. Ettling et al [42] found the amount of hydrocarbons in the
char does not necessarily indicate that accelerant had been added. However,
under careful experimental conditions, the GC plots yielded fi'om char with
accelerant in it could be distinguished from the plots of residues of wood, paper
and textiles without accelerant in them. A year later, the Research Laboratory
of Shell Oil Company [47] systemically studied C3-CO hydrocarbons in full
range motor gasolines by capillary GC. Approximately 240 peaks are observed;
180 of them had been specifically identified. This research set up a solid
foundation of petroleum-based product analysis by GC. In 1971, Chisum et a1.
[41] at the first time employed digital log electrometer instead of traditional
linear electrometer as a recording approach in arson accelerant analysis
covered four decades of signals without a range change and provided a
continuous line graph showing all of the weak and strong peaks resolved by
the column. Therefore, its chromatographic pattern provided a substantially
improved means for identifying and distinguishing hydrocarbons commonly
encountered as accelerant in arson cases. In 1976, M. H. Mach [48] utilized
a computerized Finnigan GC-MS spectrometer with methane chemical
ionization to characterize samples of simulated arson residue derived from
gasoline by distillation, evaporation, and combustion. His result showed the

more concentrated samples demonstrated the presence of higher polycyclic

30
aromatic hydrocarbons (PAHs) not seen in the original gasoline or the early

distillation residues. He believed that if these materials could be distinguished
from compounds derived from their carriers like wood, plastics, etc., routine
analytical techniques can be developed, based on the presence of these
characteristic PAHs, to determine whether or not gasoline was used as an
accelerant in a suspected arson case. In 1977, a comprehensive and
comparative study was conducted by I. C. Stone et al [49]. Headspace GC-MS,
IR, XRD, GLC, NMR and even TLC were employed, and the conclusion was all
the above instrumentations could provide complementary and corroborating
information. A study [50] of 80 crude oil GC plots which were yielded from 4
oil types by the pattern recognition technique with 13 descriptors (peak areas
of 13 particular components) showed that there are strong similarities between
un-weathered and weathered oils of a same type. However, there were also
dissimilarities present. They suggested that a more reliable method requires
a model for the weathering process. But it is common sense that the more the
weathering process is standardized, the further away from the real arson
situations the research is getting. In 1979, a study done by P.J. Loscalzo etc.
[51] studied an important issue of arson accelerant identification -- the limit
of detectibility of gasoline vapor from simulated arson residues, which was
operationally defined as the maximum time period allowed for collection and
analysis of samples in which a positive result by GLC could be obtained. Their

research measured each sample by using various combustion time and

31
collection delays after the fire was extinguished. They concluded the GLC

detectibility is 20 minutes of combustion and 162 hours of delay. In 1986, D.
C. Mann’s study [52] exhibin that all the 12 neat petroleum products of his
study collected at random were easily distinguished by a capillary GC plot
comparison method he designed. But he also admitted that when this method
was applied to general case work, it is only useful in eliminating the possibility
of common origin between samples and much less useful in determining

conclusively that two samples could have same origin.

A literature review of petroleum-based product studies shows that though
many papers deal with the detection of gasoline residues in arson cases [53-
57], only a few cover the identification of the brand, the production plant or lot
to lot individualization, or to match Sample B with Sample A. In 1988, Robert
Hirz [15] investigated individualizing of gasoline by GLC. According to the
results, identification of the refinery is possible for leaded, liquid gasoline
samples whereas lot to lot individualization is possible with liquid, slightly-
weathered or un-weathered samples. Highly weathered samples, like fire
debris from successful arson, can neither be used for identification of the

production plant nor for lot to lot individualization.

To sum up, the GC can readily distinguish types of petroleum-based products,

for example, it can differentiate gasoline from other types of flammable liquids

32
such as kerosene, turpentine, and naphtha, or differentiate various groups

among each of the above general types, with current instrumentation. It still
can not offer satisfied solution however, for the identification of petroleum-
based products by brand name by itself, or determines if two samples could

have had a common source.

4. THREE DIMENSIONAL FLUOROLIETRY.

Starting about three decades ago, criminalists have been nearly unanimous in
judging gas chromatography as the most sensitive and reliable instrumentation
for detecting and characterizing petroleum-based product residues from arson
cases. However, a new technique is becoming more well known during the
recent decade. It is fluorescence spectroscopy, especially its new generation

techniques -- three dimensional fluorescence spectroscopy.

As pointed out before, there is no single method for the individualization of an
oil. For example, the US. Coast Guard Research and Development Center
used infrared spectroscopy, gas chromatography, thin-layer chromatography,
and fluorescence spectroscopy in its basic oil identification protocol [58]. The
US. Environmental Protection Agency used gas chromatography, fluorescence

spectroscopy, and ultraviolet absorption spectroscopy [59]. However, the

33
possibility of using fluorescence spectroscopy has received increased attention

recently. By years of systematic research, both EPA and the Coast Guard
believe that fluorescence spectroscopy is singularly the most useful diagnostic
tool [59]. Virtually all of the fuels and lubricants derived from petroleum
exhibit significant fluorescence due to the presence of various types of PAHs

[60].

Along with the birth and growth of quantum mechanics, the phenomenon of
luminescence has been thoroughly studied in this century. Superficially
speaking, in normal physical conditions (for example, room temperature,
atmospheric pressure on the surface of the earth), most electrons rotating
around the nucleus of a molecule possess a series of closely spaced energy
levels with the lowest being called ground state. However, when external
energy is introduced to this ground state molecular system, for example, in the
form of optical illumination, these electrons may absorb a certain amount of
energy based upon the specific molecular structure of a specific compound, and
jump up, or in physical term "transit" to higher energy levels (excited state).
This process is called excitation. However, the excited electrons at higher
energy levels are very unstable. Spontaneous returning to their original lower
energy levels occurs, and causes these electrons to release the extra energy
they just got from excitation. This release can take place in two ways. Most

of the released energy is in the form of electromagnetic energy, the rest is in

34
the form of thermal energy. This process is called emission. If the released

electromagnetic energy is in the form of either ultraviolet, visible or infrared
light, the phenomenon is designated as luminescence. Luminescence can be
in the form of either fluorescence or phosphorescence. The former appears at
shorter wavelengths and has a faster decay time (10'12 - 10" second) while the
latter appears at longer wavelengths and has a longer decay time (10" to 10
second) [61]. Because the measurement of phosphorescence spectra require
much sophisticated instrumentation (for example, linear matrix detector,
synchronous shutters to control excitation and detection time sequence, etc.),
it has not been commercially developed. On the other hand, normal

fluorescence spectrometer has been available for about two decades.

For any one sample, normal fluorescence spectroscopy consists of two spectra:
excitation and emission. They are obtained by scanning either the emission
or excitation monochromator through the ultraviolet-visible electromagnetic
spectrum while holding the other monochromator constant, at a wavelength
where significant luminescence is expected to occur. Normal fluorescence
spectroscopy yields information about the luminescence properties of a sample
under only a single set of instrumental conditions, for example, a fixed
emission wavelength with a corresponding excitation spectrum, or a fixed
excitation wavelength with a corresponding emission spectrum. In terms of

sensitivity, normal fluorescence spectroscopy is superior than many other

35
instrumental analytical means for samples which are pure compounds and

have high quantum yield. This is why it has received wide applications in
many areas such as environmental pollution monitoring, biochemical analysis,
medical testing and so forth. In terms of selectivity, both fluorescence and
phosphorescence spectra are usually characterized by yielding several
relatively broad band peaks under the condition of room temperature, whereas
UV, IR, MS or AA can offer several dozens or more than a hundred narrow
peaks. Therefore, in the presence of complex mixtures, normal fluorescence
measurements offer limited analytical application due to at least two reasons:
(1) Strongly overlapping and interfering broad peaks are hard to interpret. (2)
Normal fluorescence is performed under only a single set of instrumental
conditions, this set of conditions could be appropriate to one component in the
mixture to get optimum excitation or emission intensity but incompatible with
other components. Unfortunately, petroleum-based products are complex
mixtures which consist of a number of different kinds of PAHs. Accordingly
normal fluorescence is very sensitive and suitable to detect the existence of
trace petroleum-based products quantitatively, or classify petroleum-based
products between different groups, but not adequate to identify whether two

petroleum—based products could have come from the same source.

In the last fifteen years, multidimensional analysis has become an important

trend in the whole realm of instrumental analysis. Multidimensional data can

36

provide chemical information as a function of more than one variable and,
hence, is especially useful for the examination of complicated mixtures.
Examples of multidimensional, or multi-parametric techniques include gas
chromatography/mass spectrometry (GC/MS) [62], gas chromatography-infiared
spectrometry (GC/IR) [63], chromatopolarography [64], secondary ion mass

spectrometry [65], and fluorometry -- excitation-emission matrix (EEM).

EEM or Three-dimensional fluorescence spectroscopy represents a great
improvement over normal fluorescence because it is capable of providing the
total luminescence of petroleum-based products instead of a single narrow
region. It retains the advantage of normal fluorescence -- high sensitivity
while overcomes its disadvantage -- low selectivity. 3-D fluorescence involves
the collection of a large number of fluorescence spectra, each taken under
different instrumental conditions, for example different excitation or emission
wavelengths. This collection of spectra is then plotted on a single graph
resulting in a complete picture of the fluorescence characteristics of the
sample, far more information than would be available fi'om a single

fluorescence plot.

In a spectrofluorescence measurement, there are three intrinsic parameters:
the excitation (EX) wavelength, the emission (EM) wavelength, and the

fluorescence intensity. They may be readily depicted in a three-dimensional

37
representation. Instrumentally, 3-D florescence can be accomplished in two

ways. One is "video fluorometry", the other is "computerized re-composition

fluorometry".

Video fluorometer approach adopts two major innovations comparing to normal
fluorescence instrument. The first is the slits of both excitation and emission
monochrometor are widened from several nm (e. g, 3 nm) to more than a
hundred nm (e.g., 250 nm). This means, first, the electromagnetic energy of
the total wavelength range which is intended to measure is projected to the
sample simultaneously. And second, the fluorescence emitted by the sample
is also transferred to the detector simultaneously. The gratings in EX and EM
monochrometors are mounted perpendicularly to each other; consequently, at
the optical plane of the detector, EX wavelength will be arranged from high to
low along one axis (say, x-axis) and EM wavelength will be arranged from high
to low along with another axis (say, y-axis). This forms a matrix at the
detector plane. Each element of the matrix represents a pair of EX-EM
wavelengths and the value of this element (z-axis) represents the fluorescence
intensity. The second innovation of the video fluorometer is that it uses a high
resolution/high sensitivity TV camera tube instead of the traditional single
element detector. By fast scanning the elements one by one in a line and line
by line in the matrix, a frame of fluorescence can be picked up by the electronic

beam in the TV camera and further processed by an on-line computer to

38

construct 3-D fluorescence in different representations. The advantage of video
fluorometry is it can get all the needed information nearly simultaneously (for
example, within 1/30 second) and makes it ready to process or printout. This
is especially important to the samples which are subject to chemical change
under long time luminescence. However, until now, this technique has still at

the stage of laboratory experimentation, and not commercially available.

This study adopts computerized re-composition fluorometry to generate 3-D
fluorescence, which does not dramatically change the monochromatic system
and the detection system. Its innovations are principally in two aspects. The
first is the mechanical automation of EX and EM scanning procedures under
computer control. The second is the adoption of a powerful computer work
station to process the data yielded by the detector under preset scanning
conditions. In the case of this study, during each EX or EM scan, the scanning
monochromator covered the same wavelength range while the other
monochromator was fixed at a certain wavelength. Between different scans of
the same wavelength range, the fixed monochrometor incrementally moves to
the next fixed wavelength, and stops at this wavelength for the next scan. By
this fashion, the fixed monochrometor will incrementally scan several dozen
times and over the entire range of luminescence. The process is automated so
that the operator need not be present for each scan. The resultant spectra are

saved by the computer and then subsequently plotted on a three-dimensional

39

Cartesian system with the scanning monochromator wavelength on the x-axis,
the fixed monochromator wavelength on the y-axis, and the fluorescence
intensity on the z-axis. The resultant plot can be presented in two ways. The
first way is so called 3-D stacked fluorescence plot. It so obtained is a profile
of all of the major areas of fluorescence of the sample. The profile is like the
figure used in cubic projective geometry. As many as four different views of
the contour plot may be obtained by generating both excitation and emission
3-D fluorescence plots and by plotting the fixed monochromator response from
either high to low or low to high wavelength. Figure 1 presents the four views
of the 3-D stacked fluorescence plots for a motor sample. The second way is
so called contour plot which looks like the isograms used in topography or in
meteorology. In this kind of plot, x- and y- axis are represent EX and EM
wavelength representatively, or vice versa. The plot consists of many non-
intersected and self-closed curves. The highest fluorescence intensity is the
highest at the center of the plot and gets lower toward the outside edges.

Figure 2 presents a 3-D contour plot of a charcoal lighter sample.

The high sensitivity, high selectivity (discriminating power),
nondestructiveness, and informativeness, coupled with its ability to directly
dealing with mixed samples, make 3-D fluorescence a very promising method

that can be recommended in forensic petroleum-based product analysis [66].

0M 3ND"

MEI! or scans: 3O
Extlmifili IML’REIIEM: 4 fit!
/ “R\

(a) ,
a
_m Exmmon

——/ H?
A.
be:
‘5 till EMSSIOH ' 410 fill

.
.'

 

 

 

 

 

 

 

 

 

 

 

 

0013:1011
mm: OF scans: at
EXCITRTIOH lltCiEhEltl: 4 ti!
IflYEflSIY?
‘\_ , ° m
F’. J._-:‘:
‘. __
é,“ ,
j \ ~ I ‘ acumen
0“) ’ \g“
88!"!
275 m1 sussxon m m

Four different views of the 3-D fluorescence stacked plots

FIG. 1
of a motor oil.

41

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

QDI3HtK
NUfltEP 0F EEnNS: 3?
EniiiiOH InCPEnEnT: S nn ' ' ".
5“ 137015; I I,
{:-
27m
(c)
MR“ :
M
H _Y —+
’7:— ' A Y 15m
200 NH EfiCITsTIOH 350 NH
0013H0!
NURSE? 0F SCfifiS: 38
EniSSIOn IHCPEnEnT: 5 an .........
utICItNII
. ‘ - .usnn
‘ Q‘ W:
A 1". \ .
fi:<‘\‘ ‘ \ 3%“
‘1: emssmu

 

“w \;'\ “Haw
“tear—W ._

:L‘M;m\-~vt :— -‘-
x\::::::::::::---v--*~r-F:7 ‘{ (k ‘

EECI?nTION

 

 

 

 

FIG. 1. Continued.

42

   

 

 

     

 

 

  
 

 

 

 

 

 

 

 
 

 

 

 

 

275.88 elission (M) 488.88
*' ‘" 273.88
a
1x
n
.4.
t
r1, +
J_L . . J 'fi -
Jun—r 1 A I. A a» V" g
(“x (69mm 4).?
l. a; 1;] ; :f‘! 2
U gm" 3.1 ‘ x" ‘ 3
v ~Thres
205.“
H.
( m 0

FIG. 2. 3-D fluorescence contour plot of a charcoal lighter sample.

43
As a matter of fact, a literature review reveals that the practical application

of fluorometry in analyzing petroleum-based products has received much
attention within the past decade. Furthermore, unlike GC studies which were
primarily focused on detecting the existence of petroleum-based products,
fluorescence studies were mainly concentrated on the individualization of
petroleum-based products. In 1977 , the US. Coast Guard Research Center
first employed fluorescence for identification [67] purpose and successfully
characterized spilled oil samples. Similarly, in 19703, crude oils have also been
subjected to fluorescence [68, 69], as have fuels, and all claimed better
discriminating power than other means of instrumental analysis. Several year
later, Kubic and his coworkers [61, 66] conducted a remarkable research of
engine oil individualization by the means of conventional, synchronous
excitation, and variable separation synchronous excitation (V SSE) fluorometry.
sixty-one neat and forty-five used automobile engine oil samples were
analyzed. The results showed very high degree of individuating ability. Only
two samples were considered to be indistinguishable. In 1985, T. M. Rossi and
I. M. Warner [70] developed an algorithm for spectral matching of excitation-
emission matrices (EEMs). Their "pattern recognition" method operates
completely in the "frequency domain after Fourier transformation of unknown
and known polynuclear aromatic (PNA) samples. In 1986, a method [71] called
constant energy synchronous luminescence spectrometry (CESLS) was reported

to be applied in environmental analysis of gasoline and crude oil samples. This

44
study pointed out that, (1) difl‘erent dilution is needed for different sample; (2)

low temperature (770K) provided better spectral resolution and enhanced
discriminating power of two very similar samples; (3) the sensitivity, selectivity
and reproducibility of CE SLS could fingerprint polycyclic aromatic
hydrocarbons (PAHs) in environmental samples. Due to lack of
instrumentation, most of the above studies can not directly yield true 3-D
fluorescence plots. Systematic studies of petroleum-based products
encountered in criminalistics by a commercial 3-D fluorometer were performed
only by J. A. Siege] [60, 72, 73, 74]. In his 1985 research, twenty-one unused
samples (ten motor oil samples and eleven lubricating oil samples) were
analyzed by 3-D fluorescence. In his 1987’s research, ten whole gasoline
sammes (including leaded and unleaded) were analyzed by 3-D fluorescence.
A comparison program was designed to determine if two measured samples
come from a common source by subtracting their 3-D plots. The results of
these two studies showed 3-D fluorescence provides much more spectral

information than is available from conventional forms of fluorescence.

The INTRODUCTION part of this paper can be summarized as the following

points:

1. In the United States, A remarkable percentage of fires are not the result

of natural or accidental causes, but the result of arson.

45

Arson is a very devastating and costly crime. Therefore, it is an area
which is worthwhile to devote research time and efi‘ort by law
enforcement as well as forensic scientists.

Accelerants are often involved in arson cases.

The laboratory analysis of accelerants play a very important role in: (1)
determining the nature of the fire, (2) case investigations by the law
enforcement agencies, (3) prosecution procedures in court.

Various types of materials could be used as accelerant.

The most common accelerants are petroleum-based products such as
gasoline- and kerosene-based materials (kerosene, fuel oils, and so no).

The detection and analysis these common accelerants have been the
subject of numerous books and articles over the past thirty to forty
years.

Up to now, among numerous kinds laboratory analysis techniques, gas
chromatography is regarded as the most powerful and useful method.

Fluorescence spectrometry, especially three-dimensional fluorescence
spectrometry is a newly emerging technique in accelerant analysis, but

it is still in the stage of deve10ping its potentials.

of

pr

dis
3112

gas

PART II. RESEARCH OBJECTIVES.

By critically studying the previous research in accelerant analysis mentioned
in the introduction, some inferences can be drawn. First, although a variety
of studies had dealt with the analysis of arson accelerant by either GO or
fluorescence, one class of hydrocarbons have not been given much attention,
which is the so called mid-range petroleum products. This group mainly
consists of paint thinners, charcoal lighters and certain synthetic turpentine
products. In terms of possible arson accelerants, the products in this group
have the following characteristics: (1) They are commercially available. (2)
They are commonly found around the home. (3) They are highly combustible.
(4) A significant percentage of arson cases showed that midrange petroleum
products were involved as the accelerant. Therefore, it is somewhat surprising
that these materials have not been the subject of much research into their
classification and methods of detection and analysis. A detailed examination
of the literature turns up little in the way of research specific to this class of
products, and the only method reported for the analysis of these materials is
gas chromatography, either with packed or capillary columns. Also, as
discussed above, the technique of 3-D fluorescence has been employed in the
analysis of several types of petroleum-based products including motor oils [60],

gasoline and lubricants [73], and, petrolatum products [74]. Second, although

46

47
in the field of accelerant analysis, GC is the most accepted technique; 3-D

fluorescence shows much potential, the comparison and evaluation between the
two techniques have not been performed. The author believes this comparison
and evaluation would be helpful for further development of accelerant analysis.
Third, the analysis of evaporated and burned accelerant samples had not been

conducted very much, especially by the means of 3-D fluorescence.

Recognizing the above three points, the present study designed to employ 3-D
fluorescence in conjunction with capillary GC to characterize mid-range

petroleum distillates under the conditions of

1. neat samples,
2. evaporated samples, and
3. burned samples.

The objectives of this study were:

1. to determine if midrange petroleum products could be classified in
groups according to their fluorescence and gas chromatographic

characteristics and to see what extent these two groups correlate,

2. to determine if the 3-D fluorescence spectrum and the gas

chromatogram of a given midrange petroleum product are unique with

48
respect to the other members of its class,

to determine if the 3-D fluorescence spectrum and the GC chromatogram
of an unknown sample of a midrange petroleum product can be matched

to the particular source from which it came,

to determine if samples which have been partially or totally evaporated

can be matched to a particular source, and

to determine the effects of combustion upon the results of 3-D

fluorescence and GC of selected midrange petroleum distillates.

PART III. MATERIALS AND
INSTRUMENTATION CONDITIONS.

3.1. SAMPLES.

Twenty-two midrange petroleum-based products were selected for the study.
Eight of them were bought from local commercial sources during the summer
of 1987. The rest were obtained from the Federal Bureau of Alcohol, Tobacco
and Firearms National Laboratory, Rockville, Maryland, they are all available
mid-range petroleum products samples which the Laboratory had collected on
a national base at that time. The code, brand name and manufacturer of each
of these samples is listed in Table 1. The code number is created by the

author for the sake of research convenience.

TABLE 1. Samples in the Study.

 

 

SERIES# CODE# BRAND NAME MANUFACIURER
Paint Thinners:

1 PT01 Sunnyside Sunnyside Co.

2 PT02 Tru-test Cotter & Company

3 PT03 Parks Parks Co.

4 PT04 Sunnyside Sunnyside Co.

(odorless)
5 PT05 UGL Raizoff ---
6 PT06 NASCA «-

49

The
Wu
Pm]

tode

50
(TABLE 1. Continued.)

 

SERIES# CODE# BRAND NAME MANUFACTURER

 

Charcoal Lighters:

7 CL01 Gulf Lite Gulf Oil Co.
8 CL02 Poly-start Midwest Polychem
9 CL03 Northland Linwo Industries Ltd
10 CL04 Wizard Boyle-midway Inc.
11 CL05 Farmlite ---
12 CL06 Boron «-
13 CL07 Sparks «-
14 CL08 Chefs Choice ---
15 CL09 Gulf Lite «-

(another batch from CLO 1)

Mineral Spirits and Miscellaneous Others:

 

16 MSOl Varsol ---
17 M802 Parks 100% ---
18 M803 Hechinger ---
19 TT01 Sears Tirpolene «-
20 LT01 Sears Lacquer Thinner m
21 ST01 Parks Shellac Thinner ---
22 GT01 GUM Turpentine ---
3.2. Fluorescence Spectrometer and Its Conditions:

The fluorimeter used was a Perkin-Elmer MPF-66 Fluorescence Spectrometer
equipped with a Perkin-Elmer Model 7300 Data Station and PR-310
Printer/Plotter. In the initial step of the study, each sample was prescanned

to determine:

tin

Ta]

Spe

(We

Seem

51

(1) the maximum excitation and emission wavelengths,
(2) the suitable excitation and emission scanning range,
(3) the suitable excitation and emission slit widths, and

(4) the suitable scanning speed.

For a particular sample, some instrumental conditions such as scanning range
(for both EX and EM) can be automatically optimized by the function of
prescan. Therefore, by investigating the prescaned EX and EM plots of all the
22 samples, weighing the pros and cons were needed at least in the following

three fundamental aspects.

First, the chosen EX and EM scanning ranges of this study had to be a
compromise among the samples. It means, the range should be inclusive in
terms of covering as many major peaks of each sample as possible; at the same
time, it should be exclusive in terms of eliminating as much no-excitation-
range and no-fluorescence-range as possible. Therefore, talking about a
specific sample spectrum, some minor peaks could be excluded, or some flat
range could be included due to the consideration of the features of other
samples. But the maximum excitation and emission wavelengths of any of the

twenty-two samples should be included.

Second, the chosen end-wavelength of the excitation scanning range and start-

52
wavelength of the emission scanning range had to be a compromise: (1) The

scanning ranges of excitation and emission cannot be overlapped, i.e., the end-
wavelength of the EX spectra should be lower than the start-wavelength of EM
spectra. Otherwise, within the overlapping range, the scattering
electromagnetic energy of excitation caused by the sample liquid scattering will
directly pass through the emission monochromator, reach the detector and
provoke a very strong interfering response (false signal). (2) The comparison
of spectra of different samples was a major part of this study. Consequently,
it is appropriate for all samples to have the same EX as well as EM scanning
range. (3) Different samples are likely to have different excitation and
emission spectral range. So, it was necessary to take the prescan results of all
samples into consideration and create a trade-off which sacrifices both less EX
and EM spectra of the samples as whole, but may not sacrifice the lowest EX
and EM spectra for a particular sample. In other words, for a particular
sample, there could be a better choice of the end-wavelength of EX and start-
wavelength of EM which could gain a more integrated EX and EM spectra for
that sample. But when considering the whole set of the twenty-two samples,
that decision could cut off some meaningful parts of either EX or EM spectra

of more than one other samples.

Third, the chosen width of either EX or EM slit of monochromator of the

fluorometer had to be a compromise of sensitivity and selectivity. This means,

53

if the slit gets narrower, the selectivity of measurement will increase. In other
words, the ability to distinguish two sharp and near peaks in a spectrum will
be increased. But at the same time, the electromagnetic energy which could
pass through the optical system and be received by the detector will decrease.
In other words, the ability to catch weak fluorescent peaks (which could be
crucial to differentiate two similar spectra) will be diminished. By the same
token, widening the slits will decrease the ability to detect the narrow and
near peaks but increase the ability to detect the weak energy peaks.
Fortunately, unlike an atomic emission spectrum, which usually occurs at high
temperature and gas state conditions, and is characterized by having many
sets of very narrow spectral peaks (lines), our samples belong to the category
of molecular spectra, which usually occur in low temperature (for example,
room temperature or lower) and in liquid or solid states, and is characterized
as having relatively broad spectral peaks or hands. Accordingly, it is possible
to choose relatively wide slits to guarantee the detection of weak peaks. At the

same time, it is unlikely to blur the spectral details.

Based on the above considerations, the instrumental conditions for the

analysis, determined empirically in the initial stage were listed in TABLE 2.

54

TABLE 2. The Instrumental Conditions of
MPF-66 Spectrofluorimeter.

 

 

Excitation (EX) scanning range 205 to 273 nm
Emission (EM) scanning range 275 to 400 nm
EX slit 5 nm
EM slit 5 nm
Scan speed 120 nm/min
Numbers of scans 35
Signal response AUTO
Emission filter OPEN
Plotting choice (1) Stacked (a) EM: low-high
high-low
(b) EX: low-high
high-low
(2) Contour THRESHOLD: 8 %
3.3. Gas Chromatograph and Its Conditions.

All gas chromatography analyses were performed on a Varian Model 3300
equipped with a Model 601 Data Station and a Hewlett-Packard Think Jet
printer. The capillary column was a J &W DB-l, 30-M, 0.25-micron coating.
The oven was temperature programmed from 50° C to 250° C. The detailed
temperature programming conditions were listed in TABLE 3. Before this
study, the experience of Dr. Jay A. Siegel’s forensic sciences laboratory had
indicated this combination of column and temperature conditions allowed good
separation, low baseline, and high signal/noise ratio for hydrocarbons while

still permitting a short analysis time (apprordmately 25 min per sample) for

55

mid-range petroleum products. The chart speed (1cm/min) used was a trade-

ofl' of adequate peak separation and sufficient peak height representation for

easier visual evaluation of the peak heights and shapes.

TABLE 3. The Temperature Programming Conditions of
Varian Model 3300 Gas Chromatograph.

 

Initial column temperature
Initial hold time

Final column temperature
Programming rate

Final hold time

Injector and detector temperature
Detector temperature

Run time

Injection size

50° C

4 min

200° C

10 deg/min
6 min

250° C

250° C

25 min

0.5 to 2 uL

 

PART IV. PROCEDURES AND RESULTS.

This study is consist of five stages. They are: neat sample grouping by both GC
and 3-D fluorescence, 3-D fluorescence concentration study, blend test to
compare the discriminating power of GC and 3-D fluorescence, evaporated

sample study and burned sample study.

1. STAGE 1. NEAT SAMPLE GROUPING.

"Neat Sample" in this study is defined as (1) the sample which was directly
obtained from either retail market or the supplier maintained above; (2) they
were well protected by glass or metal containers at room temperature after
they were received by Jay A. Siegel’s laboratory; (3) they were not

contaminated, evaporated or burned before this stage of the study.

The reasons to conduct the "neat sample grouping" were:

1. From the instrumental point of view, most previous studies utilized GC
and 3-D fluorescence to identify arson accelerant separately, and both
claimed substantial progress. However, few studies had been done to
compare the discriminating power of arson accelerant by GC and 3-D

fluorescence. As the purpose here is to compare the two instruments,

56

57
it is necessary to eliminate other factors which could influence the

results as much as possible. A major set of influential factors were
"sample variation" by their environment, such as evaporation,
weathering, contamination, burning. Only in this way, can the variable
of interest -- the difference of the discriminating power between the two
means be explored explicitly and undoubtedly. In this regard, neat

samples were more suitable than evaporated or bumed samples.

2. From a practical point of view, in certain incendiary fires, some of the
petroleum-based products used to accelerate the fire may be recovered
in unaltered or slightly altered form. For instance, the remains of a
Molotov cocktail are recovered partially intact soon after a fire, and it
constitutes essential crime scene evidence especially if a suspect is
apprehended and has in his possession a container that contains some
petroleum-based product which is suspected to be the source of the
material found from the crime scene. This kind of sample is similar to

the condition of "neat sample".

Neat sample grouping was conducted in two steps. The first step was grouping
the twenty-two samples by 3-D spectral plots. The second step was grouping

them by gas chromatography.

58
In the first step, each of the 22 samples was diluted into the solutions with

three concentrations: 1000ppm, 100ppm, and 10ppm (v/v) by spectrograde
cyclohexane (Burdick & Jackson 00.). T. A. Kubic etc. [61] found that
impurities in the solvent can frequently cause false sample fluorescence or can
yield spectra that coincide with those of petroleum-based product samples.
These effects can only be minimized if high-quath spectral grade solvents or
a suitable purification system is used. Therefore, this study chose cyclohexane
as the solvent from a possible solvent pool by experimentally determining it
has lower fluorescence than others in the wavelength range which were
intended to adopt for sample measurement, and chose the highest quality
cyclohexane which could be found on the market. Under the experimental
conditions of this study, the background fluorescence of cyclohexane was
examined and determined that it is not strong enough to contribute any
significant distortion to the sample spectra, and thus did not interfere in the

analytical results.

Previous experimental experience with fluorescence of petroleum-based
products has been that they are likely prone to leave some residues behind in
the glassware used in making up the solutions. To guard against any chance
of this occurring and to minimize the chance of any residual fluorescence fiom
such residues, two procedures were employed. First, glassware was only used

once before a carefully designed and performed washing process. After

59
preparing a solution for each concentration of each sample, a specific washing

process of all glassware was employed: cyclohexane followed by acetone,
distilled water, concentrated nitric acid, and then several times by distilled
water again. After the final wash with distilled water, the glassware was
dried in an oven at 60° C. By adopting this process, the residues left fi'om
previous sample preparation or fi'om the same sample but different
concentration preparation was eliminated enough to the degree which had no
influence on later measurement. Second, a "purity test" was conducted after
each three runs. This means a background measurement was performed. The
"background measurement" is the measurement which only measure the
solvent (i.e., cyclohexane) without solute (i.e.,mid-range petroleum-based
products) under the same instrumental conditions of sample measurement. If
the resulting 3-D spectrum by purity test turned out abnormal -- different from
the standard background spectrum which was obtained at the beginning of this
step under restricted conditions and showed no residual fluorescence, the above
glassware washing process and the whole sample measurements after the last
qualified purity test would be redone more carefully, until no residue inference
could be observed. This measure was very time consuming, but the author
utilized it throughout all stages of this study in order to ensure the reliability
of the work.

When it was necessary to compare 3-D fluorescence plots of measured samples

60
in order to determine if they belong to the same group (i.e., they were similar)

in this stage of the study, or if they came from the same source (i.e., they were

identical) in the study of later stages, two methods were available.

The first was direct "visual comparison". By this method, the number of peak
regions, the maximal EX and EM wavelengths, the relative intensities of major
peaks, and the general profile appearances were examined. Though this
method could be appraised as somewhat subjective, the 3-D fluorescence plots
were so informative and detailed enough to permit an accurate logical
judgement, i.e., yes-or-no judgement by an experienced researcher. The other
method was a computerized "subtraction algorithm". Software developed by
Jay A. Siegel [73] provided an algorithm whereby two 3-D fluorescence plots
(not limited to fluorescence plots) can be subtracted spectrum-by-spectrum and
the resulting difference plot can be displayed. Strictly speaking, because the
fluorometer used was extremely sensitive, even two samples which had the
same source would yield some differences caused by the instrumental
fluctuation or any random errors introduced during the process of sample
preparation. Therefore, the subtracted plot usually did not consist of several
dozens of parallel straight lines, instead, some non-zero fluorescence always
appears. However, the intensity of the peaks which were subtracted fi'om two
samples having a common source would be much lower than those having

different sources. More importantly, the software can provide Pearson’s Q

61
algorithm to calculate the peak differences between the samples. In those

cases where the samples had a common source, the Q value was typically
above 0.9 (1.0 representing perfect correlation). In most of the samples which
did not have a common source, Q was less than 0.9. In theory, this method is
more objective and accurate than visual comparison. But two concerns exist
for adopting the computerized subtraction algorithm. First, a subjective
judgement if the subtracted plot of two comparing samples was still needed.
Second, if totally turned to objective judgement, i.e., totally relied on the
critical value (Pearson’s Q is greater or less than 0.9) determination, some
practical considerations must be taken into account. Like in many fields of
experimental science, during the experimental procedure of yielding 3-D
fluorescence plots, a lot of manipulable and variable factors could influence the
results. For instance, the sample were all measured at very low
concentrations. The preparation of these concentrations was done by hand and
by using relatively coarse volumetric flasks. Therefore, concentration
fluctuation was inevitable. It is needed to point out that, unlike with some
other instrumental approaches, the variation of sample concentrations will not
cause the fluorescent intensity of different components (different peaks in plot)
to change proportionally, but will usually change the ratios of peak intensities.
This change will definitely affect the Q value to some degree. At the time of
this study, the algorithm had not yet fully tested to determine the general

"rule" can be relied upon in all cases, and in our particular experimental

22

Pa!

62

conditions. Therefore, the author decided to choose "visual comparison" in all

stages of this study.

TABLE 4. The Grouping Result of 3-D Fluorescence
Plots.

 

Group 1 2 3 4 5

 

Sample PT01 PT03 PT05 CL09 ST01
Code PT02 M802 LT01 (1000ppm)
PT04 CL01
(100ppm) (1000ppm)
PT06 CL05
(100ppm) (100ppm)
CL02 CL06
(1000ppm)
CL03
CL04
(100ppm)
CL07
CL08
M801
M803
TT01
GT01

 

By the means of cautious visual comparison, the 3-D fluorescence plots of the
22 samples were classified into one of the five groups. The results are listed
in TABLE 4. In the table, if the concentrations of a sample are not noted in

parentheses, they were measured at 10 ppm. At this stage of the study,

63

though each sample was measured at least in three concentrations (1000 ppm,
100ppm, 10ppm) in order to get more understanding of 3-D fluorescence
spectra of mid-range petroleum products, the optimum concentration of
grouping was chosen when the ratio between the scatter peak and the most
intense fluorescence peak intensities were approximately constant. This
condition implies that the efi'ects of self-absorption among different components
of a sample had been diminished at the lowest level. Accordingly, the
fluorescence feature of the sample had been undistortedly exhibited. Figure
3a through e are 3-D fluorescence stacked plot spectra of a representative of

each group.

The second step of STAGE I is grouping the twenty-two neat samples by their
gas chromatograms. In this step, as well as in all the following stages of this
study, all samples for GO measurement were original (un-diluted). It means
that they were directly drawn from their original containers by 10 ul syringe
(Namilton Co.). The injection volume was 1 ul for first run. If the
chromatograms were over-scaled or too weak, the initial plot attenuation
and/or the injecting volume of that sample would be adjusted gradually until
a suitable chromatogram was obtained. Likewise, "visual comparison" was also
adopted as the grouping criterion. The results of GC grouping showed the
twenty-two samples fell into nine groups and these presented in TABLE 5. In

the table, Group 2 was empty intentionally so that fluorescence spectra and

64
gas chromatograms which fell into corresponding groups were numbered the

same. Figure 4 a though h show representative chromatogram for each of the

eight groups.

TABLE 5. The Grouping Results of
Chromatogram. -

 

Group 1 2 3 4 5 6 7 8 9

 

Sample PT01 PT05 CLO5 ST01 PTO4 CL09 GT01 CL01
Code PT02 LT01 CL06

PT03
PT06
CL02
CL03
CL04
CL07
CL08
M801
M802
M803
TT01

 

65
B 219.: n 167.688 a 234.0

   

 

 

 

 

 

 

 

I.
x 3
a
I
‘ J
I
‘t
(a) r
. ' . . 213.99
175-“ eIISSIon (Isl) 488.88
a 218.8 m 355.388 a 236.8
3
4
I
3
I
a
(b) .

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1'

id.

 

275.88 EIISSIOfl (m) (88.88

FIG. 3. 3-D stacked emission plot of representative members of each of
the five groups in which the midrange products are classified: (a)
Group 1, PT01; (b) Group 2, M802; (c) Group 3, PT05; ((1) Group
4, CL06; (e) Group 5, ST01.

(c)

(d)

FIG. 3.

66

.A... g suu+ug

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Continued.

 

A+gnub+u1

 

(e)

FIG. 3.

 

67

»a 255.8 a 43.888 2.35.8

 

 

Continued.

 

A... . .I0.l-..°.l|

 

68

y 9 Oil

 

 

 

 

' .3.
— 9 ’9‘
£06
9.
C 6..
C 0"
—' grit “’

 

 

 

& . ...

 

 

"' :EEEEEEL-oo In!

a"
r "in

#-'—$ "l: 0n

_. u u:

f“

dry—‘1‘- n (N
F” 9:3: cu

'la’lgr

1
4
fit: it:

t n u
__ l— ll 9!)

 

(a)
FIG. 4. Capillary GC plots of representative members of each of the eight
groups in which the midrange products are classified: (a) Group
1, M802; (b) Group 3, PT05; (c) Group 4, CL06; ((1) Group 5, 8T01;

(e) Group 6, PT04; (f) Group 7, CL09; (g) Group 8, GT01; (h)
Group 9, CL01.

:83:

69

 

 

I "I

l U!

 

 

 

u an
t a: an

 

(b)

FIG. 4. Continued.

FIG- 4.

“n

r.
0 .‘
Q
Q C
O to
”'0. u C
U~
O
U
‘0
a
U

 

70

 

 

IO 38!

IJ’IQI

 

 

 

I, 153

Continued.

(C)

ll ’18

71

Cflflll S'EFO O 0 Cl!!!-

 

 

Atttl: 2“ “no: 5! I Ila/nu
c ., . "v E, an

 

 

(d)
FIG. 4. Continued.

72

 

 

 

 

 

 

 

 

 

 

 

u
“I.
N
.-
OH

 

(e)
FIG. 4. Continued.

FIG. 4.

00'.

 

73

 

 

 

 

 

6 IS.
r 2

 

 

 

(0

Continued.

74

r In.
I cor

LA ‘“

 

_ l ”l
r us

3 :u

_ 1 tn
1) 0 m

 

 

 

 

_ 09)
‘ C 09.
_.4 )2L

 

O ‘0.

(3)
FIG. 4. Continued.

FIG. 4.

75

 

 

I I1.

 

 

 

 

—a: v ”'
-_ v [’3

 

v S”
. S

(h)

Continued.

76
When the two sets of groups are compared, it is found that many of the

samples fall into the same fluorescence and GC group. These are represented

in TABLE 6.

TABLE 6. The Samples Which Fall Into Same Groups
By 3-D Fluorescence and GC Plots.

 

Group 1 2 ' 3 4 5

 

Sample PT01 PT05 CL05 ST01
Code PT02 LT01 CL06

PT06

CL02

CL03

CL04

CL07

CL08

M801

M803

TT01

 

TABLE 6 shows that, at the concentrations chosen, 16 out of the 22 samples
(about 72.7%) can be grouped the same way by both 3-D fluorescence and GC
plots, though the mechanism of the two instruments and the nature of the
features the two approaches are totally different. Thus, merely by running the
GC and 3-D fluorescence, one can immediately identify the samples that do not
fall into the corresponding groupings (assuming that the samples came from

the pool of 22 used here).

77

By comparing TABLE 4, TABLE 5 and TABLE 6, some inferences can be
deduced: (1) though the physical and chemical characteristics targeted and
instrumental mechanism employed by 3-D fluorescence and GC are totally
different, both of them yielded a biggest group —- Group 1, which occupied
about halfof the total samples of this study. (2) the discriminability of the two
approaches are mutually compensatory. For example, 3-D fluorescence could
distinguish PT03 and M802 from the same GC group (Group 2); on the other
hand, GC could distinguish PT04 and GT01 from a same 3-D fluorescence
group (Group 3), and distinguish CL09 and CL01 from a same 3-D fluorescence
group (Group 4). Therefore, solely based on the results of this stage, there was

no basis to say which of the two approaches has higher discriminatory power.

In order to test reliability and repeatability of 3-D fluorescence, under the
exactly same experimental conditions described above, each sample ran twice
on GC and three times on the fluorimeter. After careful examination of more

than a hundred graphs, some tentative conclusions can be reached.

First, in each gas chromatogram of mid-range petroleum products, there were
usually more than a hundred peaks. If taking retention time, peak area of all
these peaks, and their relative ratios of peak intensities between peaks into

consideration, even by visual examination, it was obvious that these

 

78

parameters vary not only fiom one sample to another, but also from one run
to another run of a same sample under the conditions researchers could
control. In other words, the chromatograms of difi‘erent runs of one sample
could show some easy-to-find data or visual differences. Hence, the data or
profile fluctuations prohibit us to firmly ascertain if the difl’erences between
two chromatogram indeed mean they are the results from two difl‘erent
samples. This was why the author had to put the samples into a group though
there were always some chromatogramic differences among any two GC plots
within any group. However, our judgement was not based on retention times
and relative intensity ratios of all peaks, but based on some so called "intrinsic
peaks" and their relative intensities as well as the profile configuration of a GC
plot. These characteristics were stable and repeatable, and were the

foundation for the grouping of samples.

Second, contrary to the considerable fluctuation of gas chromatograms,
repeated run of 3-D fluorescence of a same sample revealed that under
carefully controlled experimental conditions, the resultant 3-D fluorescence
plots of one sample were stable. In addition, under the chosen concentration
conditions presented in TABLE 4, the differences of profile configurations,
maximum fluorescence intensifies, maximum EX and EM wavelength positions
of difl‘erent samples among any 3-D fluorescence group, though maybe quite

subtle, were stable and repeatable by different runs. 80, these differences

79
could be used to distinguish samples within a 3-D fluorescence group. By this

further within-group examination, about two thirds of the samples in the same
3-D fluorescence groups could be readily differentiated, and therefore, one

could conclude that they came from different sources.

Third, when conducting 3-D fluorescence visual comparison within a group,
contour plots possessed higher discriminating power than stacked plots. This
can be explained by the fact that in stacked 3-D fluorescence plots, some
profile configuration features are always blocked by the higher and/or broader
curves in front of them. Therefore, from any single view (either EX or EM,
either hi-lo or lo-hi), i.e., from any single stacked 3-D fluorescence plot, some
set of spectra -- maybe a crucial set of spectra for distinguishing similar
samples -- could be blocked by other higher and/or wider spectra in front of
them. Though the data or information of this set of spectra had been detected
by the fluorimeter and stored in the computer, they could not be fully
displayed by a stacked plot. Contour plots, on the other hand, create a x-y
coordinate for each fluorescence sample data, within which, a closed curve
represent a same intensity (isogradient), and a pair of x and y values of any
point on an isogradient represent the corresponding EX and EM wavelengths.
In this way, a single contour plot, in theory, losses less information than a

single stacked plot.

 

 

80
The results of STAGE I can be summarized as: GC grouped the 22 samples

into 8 groups, 3-D fluorescence grouped them into 5 groups. It was dificult to
identify most samples within a group by GC plots due to their higher
complexity and less repeatability; two thirds of the samples within groups by

3-D fluorescence plots could be reliably distinguished.

2. STAGE 1]. CONCENTRATION STUDY.

It is well-known that the intensifies and shapes of oil fluorescence spectra
depend significantly on concentration [75]. Ifthe concentration is too high, the
inner filter effect (self-absorption) could suppress some strong fluorescence of
certain component(s) in the sample; if the concentration is too low, some weak
fluorescence of certain component(s) could not be received by the detector of
fluorimeter [76]. Accordingly, it is reasonable to assume that if a sample is a
mixture, the fluorescent feature at a specific concentration mainly reflects the
features of certain components in the sample; the fluorescent features at
different concentrations could probably reflect different components in the
sample. This presumption infers that comparing 3-D fluorescence at difl‘erent

concentrations could increase the discriminating power of the technique.

 

81
In STAGE 1, there were still approximately one third of the samples within

either of the five groups which could not be distinguished by 3-D fluorescence,
because they were too similar at the concentration listed in TABLE 4. Stage
II of this study was an attempt to enhance the discriminability of 3-D
fluorescence by running samples at different concentrations. This stage was

performed in two steps.

The first step was to ascertain the variability and stability of 3-D fluorescence
yielded from mid-range petroleum product samples at different concentrations.
It is well known that in theory, during a sample preparation procedure,
"experimental error" always exists. In the present study, for example, sample
dilutions were made by, first, drawing the original sample by a syringe; then,
diluting it by a set of glass graduated cylinders and measuring flasks. All
these procedures were done by hand. Therefore, the degree of precision was
objectively limited by the accuracy of graduation of the cylinders or flasks
available on the one hand; and also subjectively limited by technical fluency
and psychological state of the operator on the other hand. According to a
theory of scientific experimentation, results are always subject to systematic
errors, (e.g., caused by the deficiency of experimental tools) as well as random
errors (e.g., caused by operator’s errors in sample preparation). Because

fluorometry is very sensitive technique, and mid-range petroleum products are

82

very complex mixtures, it is necessary to evaluate the interference of

experimental errors before conducting the actual concentration study.

By doing this error evaluation, two samples were randomly chosen fi-om Group
1 and Group 4 (see TABLE 4). They were: PT02 and CL05. Each of the two
samples was measured at various concentrations. They are listed in TABLE
7.

TABLE 7. The Concentrations Used to Test the Variability and
Stability of 3-D fluorescence at Different Concentration.

 

Concentration 1: Original (without any dilution)
Concentration 2: 2000 ppm
Concentration 3: 1000 ppm
Concentration 4: 500 ppm
Concentration 5: 100 ppm
Concentration 6: 50 ppm
Concentration 7: 10 ppm
Concentration 8: 5 ppm
Concentration 9: 1 ppm

 

In this step, all concentrations of each sample were measured two times. This
means, for any of the 36 measurements (i.e., 2 samples x 9 concentrations of
each sample x 2 measurements of each concentration = 36 runs), each
preparation began right from a totally separate solution. In other words, each
of the two samples were run two times at all nine concentrations from two sets

of independently prepared solutions.

83
The results of this test displayed, first, one sample at different concentrations

showed different fluorescence features in terms of the 3-D fluorescence
patterns, the relative maximum intensities of EX or EM wavelengths, the
value of maximum EX or EM wavelengths, etc.. Second, all the above features
of two separately prepared runs of one sample at any of the nine
concentrations showed no substantial distinctions. This indicated that either
the systematic or random errors, or objective or subjective errors were not
significant enough to cause interference to the 3-D fluorescence results from
the available measurementation. The characteristics of 3-D fluorescence plots

under different concentrations were stable, repeatable and therefore reliable.

The second step was an actual concentration study. In this step, six samples
were utilized. They are listed in TABLE 8.

TABLE 8. The Samples Used in Concentration Study.

 

 

3-D Group 1 2 3 4
Sample CLO3 PT03 LT01 CL05
Code PT01

PT02

 

In this step of STAGE II, because of an extremely strong self-absorption effect

84

at very high concentrations and too weak fluorescence intensifies at very low
concentrafions, three concentrafions used in the last step were excluded. They
were: original (undiluted), 2000ppm, and 1ppm. So, for each sample, 3-D
fluorescence measurements were taken at six concentrafions: 1000ppm,
500ppm, 100ppm, 50ppm, 10ppm, and 5ppm. The stacked (both EM lo-hi and
EX hi-lo) and contour plots were obtained for each sample at each
concentrafion. The results of this study showed that the spectra for a single
sample at different concentrations differed markedly. This can be seen from
Figure 5a through f for the stacked plot spectra of a sample at different
concentrafions. Another more significant outcome was, when the three sets of
plots in 3-D Group 1 were examined, it could easily be seen that at the
concentrafion of 10ppm, all features of CL03, PT01 and PT02 were quite alike
and could not be disfinguished. But if examining them at other concentrafions,
for example, in this case, at 100ppm, the features of their stacked and contour
plots showed very substanfially differences (see FIG. 6). This result shows
that a new approach which has not pursued in previous 3-D fluorescence
studies, that is, when the task is to determine if two samples (not limited to
mid-range petroleum products) coming from the same source, in the situation
when it is difficult to disfinguish two samples from 3-D fluorescence plots at
one concentrafion as the results in STAGE I demonstrated, obtaining spectra
at other concentration(s) could substanfially increase the discriminafing power

of the tests.

£355
a 367.8 E 17.688 a 282.8

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

    

 

3'
+
I
r
0
I
r
(a) ‘E “M“
51:.
-- o
: I
s
_s
l
r
e
" . . 275.88
285.88 erItatIon (III) 273.88
I 267.8 I 14.288 I 288.8
3'
-t
o
3
I
+
r
-—-—- 888.88
(b)
. J 275.00
285.88 exCItatIon (m) 273.88
FIG. 5. 3-D fluorescence stacked plots of sample PT03 at different

concentrafions: (a) 1000 ppm; (b) 500 ppm; (c) 100 ppm; (d) 50
ppm; (e) 10 ppm; (0 5 ppm.

86

93:51:: . -. - I 2.1:..e 5.922 B 287.8

 

i

|

I

, i

(c) l!

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

285.88 exc::at:on {r : 2.:.88

I :22.2 ”.288 287.8

 

 

1
I
3
|
‘8
t
(d)
488.88

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

285.88 excnatcon (n)

FIG. 5. Confinued.

87

a :m I team a 22m

 

.4... Q 30.0-...Lc;

(e)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(6) 488.88
— ‘ -— 275.88
285.88 excitatIon (nu) 273.88
a 219.0 E 162.888 3 288.8
3‘
+
o
3
II
+
(r) r
(0 //-’/_Q\\\\ a- m- .__,.._,_,._..__--._.._. «r. W“... 49.9.39,
\g.
/
./
_,,./
285.88 echtatIon (nu) 273.88

 

FIG. 5. Confinued.

88
3. STAGE III. BLIND TEST TO COMPARE TWO SAMPLES.

Scientific analysis can be divided into two major branches. One is so called the
"rafional approach", which is often pursued to answer quesfions like: "What
are they?" "What is the value of a certain parameter this substance has?"
"What components are in it?" "What proporfions of each component?" In
general, conclusions of this kind of approach usually comes from: (1)
quanfitative observafions, and (2) objective measurement done by instruments.
In criminalistic, the determinafion of refraction index of glass, and blood-
alcohol analysis are examples of this branch. Tradifionally, instrumental

analysis in chemistry as well as in criminalisfic belong to this category.

Another branch of scienfific analysis is the so called "phenomenal approach",
which is often pursued to answer quesfions like: "Do two or a set of things
come from a same source?" "Is object A related to object B in terms of certain
features?" Conclusions of this kind of analysis are usually yielded from: (1)
qualitafive observafions, and (2) subjecfive judgements by experienced experts.
In forensic science, phenomenal approaches have been extensively employed
and overwhelmingly approved in fields such as personal idenfification by
fingerprints, hand—wrifing documents examinafion, toolmark comparisons,
bullet and cartridge comparisons, and forensic denfistry. Tradifionally, pattern

recognifion belongs to this category. Recently, forensic science has been

89
blurring the boundary of the two branches by combining the advantages of the

two to create a better result. Computerized fingerprint categorizafion and
idenfificafion, and voiceprint individualizafion are two examples of this trend.
The above concept may also be applied to arson cases by looking at necessity

and possibility.

In terms of necessity, in many cases involving scienfific evidence, the goal of
the analysis is not to answer the quesfion "what are they", or is not to idenfify
something (e.g., to idenfify a motor oil as Quaker State 10W-40), but to
compare a sample of unknown origin (Sample A) to one whose source is known
(Sample B). In common arson cases, the central forensic issue is to confirm if
the suspicious substance (usually in liquid state) found at an arson scene and
the substance possessed by a suspect came from the same source or one related
to the other in certain features. Prosecutors, judges, jurors, and even forensic
scienfists do not really care about the composifion of the liquids, or their
physical and chemical parameters, provided the conclusion of "sharing a same

source" could be drawn by whatever means.

In terms of possibility, as we know, all petroleum-based products are extremely
complex organic mixtures. By employing modern techniques, it is possible to
quanfitafively determine their major components and put them into more and

more detailed groups. However, in spite of its costliness and consumption of

90
time, present instrumental analysis has not reached the point to quanfitafively

determine all of the components of petroleum-based products in sufficient

degree, and to individualize any two commercial petroleum-based products
based on this rafional result.

Therefore, one of the goals of this study was to test the discriminafing power
of 3D fluorescence and GC by combining the rational and phenomenal
approaches. That means, using analytical instruments to get objecfive plots

and using subjecfive judgement to derive the "yes" or "no" conclusion.

As discussed above, the conventional task of a forensic laboratory for arson
evidence is endeavoring to match a liquid found at arson scene with a liquid
obtained from the suspect. To imitate this process, the author’s supervisor -
Dr. Jay A. Siegel prepared seven pairs of samples (14 single samples) fiom the
twenty-two samples listed in Table 1. They were all approximate 30 ml in
volume. Each was put into a small glass container, well sealed by a rubber cap
and respecfively labelled as: 1A, 13; 2A, 2B; 3A, 3B; 4A, 4B; 5A, 5B; 6A, 6B;
7A, 7B. The two samples in any of the seven pairs were chosen from same 3-D
fluorimetry group, but they may or may not came fi'om the same original
sample. In other words, the two samples in each pair were all similar, but
might or might not have been the same. The author - the examiner received

labelled containers with samples, but did not know which pairs were fi'om the

91

common source. His task was trying to determine the "sames" from the

"similars" by the methods and condifions developed in early stages.

The GC instrumental condifions used here were same as listed in TABLE 3.
The criterion to conclude the two samples in a pair were from a same source
was: all the retenfion time values of characterisfic peaks and the intensity
relafionships among these peaks for the two samples ought to be matched
(means they are same brand); otherwise, the two samples would be regarded
as coming fiom different sources. The determination of which peaks were

"characterisfic" was based on experience and other research results.

The 3-D fluorescence instrumental condifions were the same as listed in
TABLE 2. All the samples were first diluted to 10ppm, if no substanfial
differences could be found between the two plots in a pair at this
concentration, addifional measurements were taken at higher concentrations
in the sequence 50ppm, 100ppm, 200ppm, 500ppm, 1000ppm. At the
concentrafion where substantial differences between the two were confirmed,
the matching procedure stopped. The criterion used for reaching the "same
source" conclusion was: only if the pairs of spectra showed no substanfial
differences at all the six concentrations would they be declared to be of the
same source. Figure 6a through (1 demonstrated that at 10 ppm sample 7A and

7B could not be disfinguished. However, a cautious comparison at the

92
concentration of 100 ppm of these two samples showed substantial difl'erences.

Accordingly, it is reasonable to conclude 7A and 7B came fi-om different

80111088.

The results of the blind test are in TABLE 9. In this table, "difi’erent" denotes
that by the means of that instrumentafion, the resultant plots showed
substanfial difl'erences; "same" denotes by the means of that instrumentation,
the resultant plots showed no substantial differences. "Conclusion" means

what the Dr. Siegel really did about that pair of the two samples.

TABLE 9. The Comparafive Results of GC and TLS Blind Test.

 

 

Code of GC Fluorimetry
Sample Pair Judgment Judgment Actualities
1 same different different
2 different different difl‘erent
3 same same same
4 different different different
5 different different difl‘erent
6 same same(?) same(?)
7 same different different

 

93
Because Group I of either GO or 3-D fluorescence contained about half of the

22 samples, and either the G0 or 3-D fluorescence plots in Group I were very
similar, six pairs of the samples were chosen fi'om this group. The actual

composifions of the pairs of samples is given in TABLE 10.

TABLE 10. The Samples Used

 

 

In Blind Test.
Code of
Sample Pair Sample A Sample B
1 PT03 M802
2 CL03 CL08
3 PT02 PT02
4 M803 T101
5 CL05 CL06
6 CL04 (*) CL04
7 PT01 PT02

 

The results of blind test showed that GC was correct in 57.1% of the cases,

whereas 3-D fluorescence was correct 100% of the fime.

It is worthwhile to explain the meanings of two question marks appeared in
TABLE 9 and an asterisk mark appeared in TABLE 10. The GC
measurements of the pair 6A and 6B showed exactly same plots in all criteria
maintained above (see figure 7a through b). But the results of 3-D fluorescence
contour plots of the two samples at 10 ppm exhibited very substanfial

differences (see figure 8a through b). From experience and careful inspecfion

94
of 3-D fluorescence plots of figure 8a and b, the examiner speculated that the

majority components of sample 6A and 63 were same but an addifional
substance was in 6A. This guess was confirmed later by Dr. Siegel that he
intenfionally added less than 0.1 ppm of another reagent in sample 6A to test
the sensifivity and discriminafing power of the GC and 3-D fluorescence. The
result evidently exhibited that in the case where some low percentage of a
minority component exists, or some subtle differences exist among samples, 3-

D fluorescence technique could detect this.

4. STAGE IV. EVAPORATED SAMPLES.

One of the notable properfies of many petroleum-based products is their high
volafility even at room temperature. As evidence found at a crime scene, this
kind of material is usually in a well ventilated environment and may be taken
out from its original container for a long fime period (commonly, from several
minutes to several days). Accordingly, partially or totally evaporated
accelerant are often encountered in forensic scienfific labs. Though it is often
necessary to match evaporated unknown samples with whole knowns, there is
little in the literature to contribute to this quesfion. So, the purpose of this
stage was to determine what effects, if any, there would be on the 3-D

fluorescence and GC of the mid-range petroleum products when they are

95
parfially or totally evaporated. It is expected that the GC plots would, of

course, be altered significantly, losing much of their volatile fiacfions even with
parfial evaporafion. Changes in 3-D fluorescence plots are more difficult to
predict. The components in protrolem products responsible for fluorescence are
presumably the aromafic and polycyclic compounds presented in these
products. They would be expected to be among the less volatile fracfions and

would therefore not be much affected by parfial evaporafion.

Because the amounts of samples supplied by the Bureau of Alcohol, Tobacco
and Firearms Nafional Laboratory were not enough to be evaporated under the
evaporafion condifions designed for this study, only the eight samples directly
obtained from stores were employed. They are listed in the first column of
TABLE 11.

TABLE 11. The Comparafive Results of Neat Sample
And 50% Evaporated Sample Grouping.

 

 

Group Numbers Group Numbers
From Neat Sample From Evaporated
Sample Code Grouping Sample Grouping
PTO 1 1 1
PT02 1 1
PT03 1 1
PT04 6 6
CL01 9 9
CL02 1 1
CL03 1 1
CL04 1 1

 

96

275.88 elission (hr) 488.
.. 272

 

 

(a)

 

uol +u+ 13x3

 

 

 

(us-4|)

 

 

 

M
53?

 

 

275.88 , 2 emission (nu) 488.8

272

 

(b)

use | 4.-.; I :xa

 

l . -. 'I .. a 285
FIG. 6. Concentration studies. At concentrafion of 10 ppm, Sample 7A

(a) and 7B (b) con ont be disfinguished; at concentrafion of 100
ppm, Sample 7A (c) and 7B (d) can be disfinguished.

   

(WU)

 

 

 

97

275.88 elission (nu) 488.88
,3 L 273.

 

(c)

 

 

 

(mu) u0|+a+|=xa

 

 

235.I

 

275.88 elission (nl) 488.8

:::j_drj/ 273
:3

 

L

(d)

(mu) uo|¥i+|=xo

 

 

 

 

 

 

285

 

FIG. 6. Confinued.

FIG. 7.

98

NH
v l
M]

'2);_;
_.~'
4 3 19s
'r 3 242
a: 1.3u9
‘ 3.bI’U
’ 3.392
"\—
Efl E145 _
4 182-:
57::
.él-J
. ‘8

“41"
AJ—

(xq 1

fi,4.721
3)

{-fi ghgé

WM
"J8

Ul l ”L"
hm

«D- ‘11:]

‘D ‘3:

(TI

 

 

9 5.595
a“
“ s 933
ch—
§““é11?2
: .41- 1'9
“=K.s.539
s ‘45
7. s§'°9°
8m
_==:;.__.r ass
+1442
h‘::::::EE§EEEh-P.?S’

 

 

 

 

 

 

3 "52573
5’88“
993:“!
143.18?
:jf can; a
é...— :3
19‘71? '”
{94883
‘iI.srs
,q—m 11.45:
L‘ 11.l3r'j
er 11, 35
‘> 11.68;;
{-1—1 3:! .11 7‘85
1 . .153
F"' 11 953
f 12.974
§ 12 274

(a)

Concentrafion studies. The GC plots of Sample
Sample 6B (b) show no substanfial differences.

6A (a) and

99

.3.
.3

Q3.

5.5.... .

3.3.31 8.. ..I.

1.3 .

31..

..? 51.15..

 

5:6,.» 4. ??c

:4 ad
.I 7. «d
. .3 1

.haU .(

Luag
I

1

, .1349

£36

a
$361

 

 

.—

 

3h"?

”um

E.

'9524

 

 

 

.D

7.—

 

.458

18

a

no 3 l
l 4. 4 31
73 fits... QZH.

5 1......—
.H.H.—_...t. . .56
1 51 lllld-

3.4. 11. 4t... 3

S"

11

g

at
«.I
9.-

 

.9. CU
3 . 3
Owl 4;

(b)

Continued.

FIG. 7.

(a)

(b)

FIG. 8.

100

_ I 219.: I 45.2» I 288.!
llllllll

 

275.88 enission (nu) 488.88
' ' " 278.88

«.-

(a)

 

§
0
3
A
3
3
v

JW5MM(D-
285Mlllll

-NUIH.I
( oosuooa

 

 

 

IIIII
ouoon
snooa

IIIBISIEBIIIIIIIII an 253 Ill «so illzns
ilalflflli

elission (nu) 588.88
273.88

(b)

mu) uo.+-+q=x-

v ~5hreshold (Z).

2asyllllll
8
A

uuossu
< ooanoao

 

 

 

 

llll
V003
N000

Concentration studies. The 3-D fluorescence of Sample 6A (a)
and Sample 6B (b) show substantial differences.

 

101

The preparation of evaporated samples employed the following process:

(1)
(2)

(3)
(4)
(5)

(6)

(7)

Put 40 ml original sample into a 100 ml beaker.

Leave the beaker with sample until the sample evaporated to 20 ml at
room temperature (25° C). It took from about one to six hours
depending on the nature of the sample used. These samples were used
to conduct the "half evaporation" experiment.

Then let the samples further evaporate until there was no liquid left.
Put about 4 ml cyclohexane into a dried beaker, and thoroughly shake.
Wait about 5 minutes, then pour the solution into a sample cell
for 3-D fluorescence measurement.

After fluorescence measurements, further condense the rest of the
solution in the sample cell at room temperature until about 0.1 ml is left
for the GC injection.

In order to prevent possible mutual contamination between samples, all
the glassware used was subjected to the sample cleaning procedure
described in STAGE I.

For the half evaporated sample results, all of the eight GC plots were

meaningful, but their features such as overall profile, retention time of

characteristic peaks, and intensity relations among peaks were all considerably

changed from those of the neat sample results. Therefore, it was nearly

impossible to trace a half-evaporated sample back to its original samme by

102
examining the two corresponding GC plots, if there were no systematically

evaporated GC plots available as standard. However, when the eight samples
were grouped and same group numbers were assigned as did for the neat
samples, same grouping pattern appeared. This can be seen in TABLE 11.
The above results revealed that at least until half-evaporation, though
evaporation affect all GO plots significantly, it does not affect their grouping
patterns. In other words, the half-evaporation process has similar influences

to the samples in the same groups.

With the totally evaporated samples, it would appear that there was some non-
volatile residue left, but GC plots were quite non-descriptive and would not be
suitable for any identification purpose. This is different fi'om the results that
one gets with gasoline or wherein a nonvolatile residue is left after all of the
visible liquid is evaporated. This outcome can be explained that unlike
gasolines, which are used as car fuel, both charcoal lighters and paint thinners
are more thoroughly refined to decrease the amount of solid residues and
impurities. The results imply that, in real cases, if charcoal lighters or paint
thinners were used as an accelerant and had been totally evaporated before
collection of evidence, it would be hard to identify the existence of accelerant
by GC, even if the residue could be identified and collected. Figure 9a through
c show GC’s of a neat sample and the same sample at half and total

evaporation.

103

In the case of 3-D fluorescence, the spectra were obtainable for both half and
totally evaporated samples. However, characteristics such as overall profile
configuration, positions of maximum EX and EM wavelengths, intensity of EX
and EM peaks of these plots were dramatically different fi°om those of the
corresponding neat samples. Like in the case of GC plots, this means that a
heavily "weathered" sample could probably not be matched back to a neat
sample of the same materials unless the degree of evaporation was known, a

rare situation in a real case.

Unlike petroleum-based product studies done by GC, 3-D fluorescence
technique is relatively new and less often found in the literature. Therefore,
all half and totally evaporated measurements were repeated three or four
times to enhance their reliability. Their results clearly showed the following
features. First, the spectra of a given sample at both half and total
evaporation conditions were stable, i.e., the spectra were the same when each
time a given sample was evaporated to the same degree. Second, in the
situation of total evaporation, different from GC measurements, by which no
meaningful plots could be procured, 3-D fluorescence measurements could
produce meaningful and stable spectra. It could provide a possibility to
identify the existence of totally evaporated accelerant by this technique, though

further study to determine better technical conditions is needed. Figures 10a

104
through c show the 3-D fluorescence contour plots under the conditions of neat,

half evaporation and total evaporation.

5. STAGE V. BURNED SAMPLE EXPERINIENT.

It is always an aspiration for an arson investigator to show that the presence
of an accelerant could be found from totally burned fire residues, and a
common source confirmation could be established between totally burned
Sample A and unburned Sample B. There have been several studies of
analysis of fire debris in order to answer this question [51, 75, 76, 77]. The
conclusion of P. J. Loscalzo’s [75] study represents the primary opinions of

studies dealing with this issue:

Unfortunately, a controlled study can not take into account the
factors affecting the same accelerant under the conditions of a full-scale
dwelling fire. In such a fire the accelerant may be either entirely
consumed or altered in such a way that the usual analytical method
commonly employed today would fail to produce any positive results.
However, it is the common situation of real arson cases that the materials
which contain suspected burned accelerant are often submitted to forensic

laboratories. It is known that combustion is an extremely complex process,

and even the most rigorous attempts to control the conditions of a combustion

105
often lead to inconsistent results. Nevertheless, it was felt necessary to study

GC and 3-D fluorescence behavior of the mid-range petroleum products under
controlled combustion conditions to get a more complete picture of these

materials in actual fire situations.

Because a large amount of sample had to be consumed in this stage of the
study, only the same eight samples (see TABLE 11) were chosen as in the
evaporation experiment. Of each original sample, 50 ml was put into a clean
1 gal (3.8 L, 16.5 cm in diameter, and 19 cm in height) paint can with a 4.4 cm
by 20 cm piece of cheesecloth which was folded into a 7.5 cm by 7 .5 cm square.
The can was located on the platform of a fume hood. The distance between the
edge of the fiont door of the hood and the can was 6 cm. The 50 ml sample
was ignited by a match. The electric ventilation system of the hood was then
turned on, the interval between two doors of the hood was narrowed to 10 cm,
and located, the slit of the doors directly facing the burning can in order to
ensure a sufficient and complete combustion. Within 5 to 7 minutes all of the
liquid sample was burned out and the fire died out. The can was allowed to
cool to normal room temperature, and then 100 ml of cyclohexane was poured
in, and the can was shaken. After 10 minutes, the solution was filtered by
gravity and evaporated to dryness at room temperature; the residue was then
reconstituted in 3.5 ml of cyclohexane, and 3-D fluorescence measurements

were taken. Then, the residue solution fiom the fluorescence sample cell was

106
put into a 10 ml glass container, and evaporated until about 20 ul residue

solution was left for GO analysis.

By the above sample preparation, the mid—range petroleum products were
totally burned out and the residue solutions were evaporated to dryness so as
to approximate the burning process and the weathering process as closely as
possible and to use a worst-case scenario. In the case of these samples,
"burned out" meant the all original samples were consumed by combustion
under the conditions of sufficient supply of oxygen and ventilation, and
"dryness" meant that there was no visible liquid left. However, "burned out"
did not mean that there was no any solid residual left by combustion, and
"dryness" did not mean that there was no any residue left by evaporation. In
almost all cases, some solid combustion products and some nonvolatile residue
could be left, as is generally the case with products derived fi'om petroleum.
This residue gave rise to the weak fluorescence and resulted in the very weak

chromatogram.

As in the GC plots of the total evaporated samples, the plots of these burned
samples were quite non-described, and the resultant chromatogram were
neither able to prove the em'stence of accelerant, nor imply the type of
accelerant. This indicated that the combustion consumed nearly all of the

hydrocarbons. A possible explanation of this result is: on the one side, the

107
burning temperature of all samples under the conditions of the study were

higher at least than 500° C; on the other side, the highest programming
temperature of CC for detecting the burned residue was 250° C. Therefore, if
during the burning process, the temperature of more than 500° C was held
long enough (several minutes is long enough), most components with the
decomposition temperature lower the 5000 C would be split and further burned
away. So, it is no wonder that the resultant chromatogram were non-

descriptive.

It was possible to obtain 3-D fluorescence plots for all samples after the
burning process described above. But they were significantly different than
either the plots for neat samples or half evaporated or totally evaporated
samples. No configuration profile patterns, positions of maximum EX or EM
wavelengths, or intensity of maximum EX or EM wavelengths could permit
tracing back to a particular sample, or for that matter, the mid-range
petroleum products as a whole. Combustion involves extremely complicated
chemical reactions, so it is understandable that a resultant 3-D fluorescence
spectrum was dissimilar comparing with its neat or evaporated 3—D
fluorescence spectrum. In addition, when the burning was repeated on the
same sample a number of times under the same burning conditions, the 3-D
fluorescence spectra were all different, indicating that the combustion was

somewhat different each time. The combustion conditions could not even be

108
known, let alone reproducible. Hence, combusting a known sample suspected

to be involved in a fire and comparing the spectrum to that of the unknown

would probably be futile.

It is possible that, had the combustion been cut short by putting out the fire
with water or by smothering it, more of the residue of the hydrocarbon would
have remained, perhaps increasing the changes for reproducibility. There are,
however, reasons to believe that this would not be the situation in a real case.
First, introducing water to the mixture increases the chance for contamination
of the fluorescence owing to impurities in the water. Although this will be the
case for most real fires, it was not desirable to add this set of variables in this
initial study. Second, smothering the fire by cutting off oxygen would not
introduce any contamination, but at the same time, is not the way real fires
are generally extinguished. Finally, the evaporation study showed that, when
a portion of the sample is lost, the 3-D fluorescence plot changes and can no
longer be matched to the neat sample. Figure 9 shows a 3-D fluorescence

contour plot for a sample after total combustion.

FIG. 9.

 

109

7 ‘5!
7 627
u’ .‘z

 

9 315

t '23

 

I.
_ g- u on
5‘31 16

 

Evaporation studies.

I

Capillary GC plots of CL02:

unevaporated; (b) 50% evaporated; (c) 100% evaporated.

(a)

110

 

Pt I I!‘

 

 

 

 

7 7‘7

 

 

(b)
FIG. 9. Continued.

FIG. 9.

111

 

 

 

 

I .0.
.DL: rm. ( “3 J) 139
:gt- 3:! J
3 E v u 570
74 = z or)

 

 

 

 

 

 

 

(c)

Continued.

112

a 2‘12 m 35.:ea 285.8

 

255.66 en-ssucr Ini) 428.88

i «la ..: r ;' - mes

 

 

l- ' u ' ,
I. ll. _ I )l ‘
uf‘llil f :P /'
:,,~.~y rut :{I
(a) utmb-;;' / .;
:z'!‘-,.u.‘—' / , z r‘ , ,.
I E'- J\\\‘—«‘32 I", ,2 2 } {x
. . ‘ h \—‘ - p r p a l
l. {\1\|.‘\\—I— v..." d a :r:
. i \i _’ .4— dd ;‘
)3:- _’_/ '. :3 _
I H H - ll
a- g - -
r/ -——. x ,0
\ J-
at . 3
A
J
l! .
v 'Threshoid t4)-

 

 

 

28$mun"
OV
5N-

'NL‘IJU
0

III
cco
( @03NGC sec

(3)

FIG. 10. Evaporation studies. 3-D fluorescence plots of Sample LC02: (a)
unevaporated; (b) 50% evaporated; (c) 100% evaporated.

113

m 8 273.0 I 30.290 I 290.3
_

275.00 eitsszon (nu) 000.00

(MW/W (
W

 

K.

C17

u’ou) uo.+--‘

.V -Ihreshoid (2)-

 

 

 

Minimum
~nossaaucon
( oasvocasnoia

m a 263.0 E 324.490 3 363.0

 

 

 

 

 

 

 

 

m
275.00 elission (nu) 400.00
' ° I — ‘73.“
A (ya/”S F—/—w ‘ ‘
(C) \W l
_. . . rm“
xii—j"; '1
___,_/"FJ_J—o
3
r— 3
R -Threshold (2)-
zflllllllllllll
'- H 0 8 8 '1 3 \J O I O
( oolnooasnooo

FIG. 10. Continued.

114

a 232.5 a $5.299 a me

 

275.00 «mm (m) ' 400.00

 

-Threshold (2)-
Jarwml

H1
( coinctl

 

 

 

FIG. 11. 3-D fluorescence contour plot of Sample CL02 after combustion.

PART V. SUMMARY

Twenty-two mid-range petroleum-based products, including charcoal lighters,
paint thinners, and synthetic solvents, were studied by capillary GC and 3-D
fluorescence. It was found that 3-D fluorescence is far better at discriminating
among similar products than was GC, which could only put the products into
broad classes. When partially or totally evaporated, the 3-D fluorescence plots
were altered enough that they could not be matched with those of neat
samples. A similar situation exists when they are burned. Also, when burned
even under controlled conditions, these compounds did not yield consistent
fluorescence patterns. A blind test on neat samples showed that 3-D
fluorescence can be a reliable technique for determining whether or not two

neat samples were of the same brand.

This study is a continuation of Dr. Siegel’s serial studies of 3-D fluorescence,
it is also a preliminary study on mid-range petroleum-based products as
potential accelerants by the means of both gas chromatography and 3-D
fluorescence. It has its limitations like all this kind of studies. First, though
the twenty-two samples were the all that the author could get at that time, he
is still not very confident about the representativity in terms of the whole
population of mid-range petroleum products. Second, no doubt the neat sample

grouping has its theoretical meanings, however, pure liquid samples are not

115

116

frequently encountered in practical arson cases. Third, 3—D fluorescence
approach usually needs larger sample amount for measurement than gas
chromatography does, this could be limited its applications in some arson

03888 .

In spite of the limitations mentioned above, this study does have a few
important implications in its technical and legal aspects. First, in accelerant
identification of arson cases, 3-D fluorescence has some advantages in terms
of either sensitivity or discriminating power comparing to GC. Therefore, it is
worthwhile to add this technique as a routine approach in crime laboratories
to get complementary information if the financial condition permits to do so.
Second, because large percent of organic substances have the property to emit
fluorescence under certain conditions, this study could enhance the status of
3-D fluorescence not only in arson analysis, but also in other aspects of forensic
sciences as well as in instrumental analysis at large. Third, it is particular
interesting that 3-D fluorescence can identify small amount of components
under the condition of the existence of other large amount of strong fluorescent
substances, like the situation of distinguishing Sample A and B in blend test
of this study. Fourth, from the legal aspect, even the proving of the existence
(if it is not possible to identify the sample) of an accelerant is helpful to
determine the nature of the fire, accordingly, helpful to the process of further

police investigation and court prosecution.

[l]

[2]
[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

REFERENCES

Benett, G. D., Journal of Criminal Law, Criminology and Police
Science, Vol. 49, 1968, pp. 172-177.

Horn, J ., Journal of Individual Psychology, Vol. 31, 1976, pp. 205-210.

Robison, M. M., and Wagner, P. E., Fire Technology, Vol. 8, No. 4,
Nov. 1972, pp.278-290.

Boudreau, J. F., "Arson and Arson Investigation: Survey and
Assessment, "Washington, D. C., Government Printing Office, 1977 ,
p.1-5.

Inciardi, J. A., "The Adult Firesetter: A Topology," Criminology: An
Interdisciplinary Journal, Vol. 8, August 1970, pp. 145-156.

Inciardi, J. A., "Reflections On Crime," New York: Holt, Rinhart and
Winston, 1978, pp. 127-128.

Ferrall, R. T., "Arson Information: Who What Where," Law
Enforcement Bulletin, May, 1981.

DeForest, P. R., and Lee, H., "Forensic Science", 1983, McGraw-Hill,
New York.

Nicol, and Joseph, D., Fire Eng., Vol. 114, 1961, p. 550.

See [42].

Midkiff, C. R. "GC determination of Traces of Accelerant in Physical
Evidence," Journal of the Association of Official Analytical Chemists,
Vol. 55, No. 4, 1972, pp. 840-845.

Midkifi', C. R., "Arson and Explosive Investigation", Chapter 6 of
"Forensic Science Handbook, Prentice-Hall, Inc. 1982.

Zieba, J ., Forensic Science Interna., 1985, Vol. 29, No.3, pp.269.

117

[14]

[15]

[16]

[17]

[18]

[19]

[20]
[21]

[22]

[23]

[24]

[25]

[26]

[27]
[28]

[29]

[30]

[31]

Zieba, J ., Forensic Science Interna., 1986, Vol. 30, No.1, pp.45.

Robert, H., "Gasoline Brand Identification and Individualization of
gasoline lots," JFSS, Vol. 29, No. 2, 1989, pp. 91-101.

Zafiriou, O. C., Anal. Chem., 1973, Vol.45, pp. 249.

Garza, M. C. Jr., and Muth, J ., Environ. Sci. Technol., 1975, Vol.8,
pp. 249.

Dell’Acqua, R. R., Egon, J. A., and Bush, B., Environ. Sci. Technol.,
1975, Vol. 9, pp. 38.

Pym, D. G., Ray, J. E., and Whitehead, V. Analy. Chem. 1975, Vol.
47, pp.1617. ‘

Rasmussin, D. V., Anal. Chem., 1976, Vol. 48, pp.1562.
Kawahara, F. K., Environ. Sci. Technol., 1976, Vol. 10, pp. 761.

Goldberg, M. C., and Devonald, D. H., J. Res. U.S. Geol. Sur., 1973,
Vol. 1, pp. 709.

Lynch, P. F., and Brown, C. W., Environ. Sci. Technol., 1973, Vol. 7,
pp.1123.

Brown, C. W., and Ahmadiian, M., Environ. Sci. Technol., 1974, Vol.
8, pp. 669.

Brown, C. W., and Lynch, P. F., Anal. Chem., 1976, Vol. 48, pp. 191.

Kawahara, F. K., and Yang, Y. Y., Anal. Chem., 1976, Vol. 48, pp.
651.

Mattson, J. S., and Starks, S. A., Anal. Chem., 1977, Vol. 49, pp. 297.
Kawahara, F. K., J. CHromatogr. Sci. 1972, Vol. 10, pp. 629.

Duewer, D. L., and Shatzki, T. F., Anal. Chem., 1975, Vol.47, pp.
1573.

Ludens, H. R., Prog. Nuci. Eng. Anal. Chem., 1975, Vol. 12, pp. 1.

Fortier, S. H., and Eastwood, D., Anal. Chem., 1978, Vol. 50, pp. 334.

118

[32]

[33]

[34]

[35]

[36]

[37]

[38]

[39]

[40]

[41]

[42]

[43]

[44]

119
Anbar, M., and Aberth, W. H., Anal. Chem., 1974, Vol.46, pp. 59A.

Shultz, W. W., "Application of High pressure Liquid Chromatography
to the Identification of Oil Spills in the Marine Environment", 26th
Pittsburgh Conference on Analytical Chemistry and Applied
Spectroscopy, Cleveland, Ohio, March, 1975.

Saner W. A., and Fitzgerald II, G. E., Environ. Sci. Technol., 1976,
Vol. 10, pp. 893.

Oil Identification Symposium, 27th Pittsburgh Conference on
Analytical Chemistry and Applied Spectroscopy, Cleveland, Ohio,
1976. »

Frankerfield, J. W., "Weathering of Oil at Sea", AD 789, 1973.

Miller, J. W., "A Mutilparameter Oil Pollution Source Identification
System". EPA-R2-73-221, July, 1973.

Proceedings, Workshop on Pattern Recognition Applied to Oil
Identification, Coronado, Calif., November, 1976, IEEE Catalog No.
76, CH 1247-6 C.

Symposium on Oil and Other Organic Pollutants Analysis, 29th
Pittsburgh conference on Analytical Chemistry and Applied
Spectroscopy, Cleveland, Ohio, 1978.

Park, B. P., and Kirk, P. L., Microchemical Journal, Vol.6, Jan, 1962,
pp.31-38.

Chisum, W. J ., "Identification of Arson Accelerant by Gas
Chromatographic Patterns Produced by a Digital Log Electrometer",
Journal of Forensic Sciences, 1971, Vol. 43, pp.280-291.

Ettling, B. V., and Adams, M. F., Journal of Forensic Science, Vol. 13,
No. 1, Jan, 1968, pp. 76-79.

Cain, P. M., "Comparison of Kerosene Using Capillary Column Gas
Liquid Chromatography," Journal of Forensic Science Society, Vol. 15,
1975, pp. 301-308.

Dell’Acqua, R., "Identification of Gasoline Contamination of Ground-
water by Gas Chromatography," Journal of Chromatography, Vol 128,
197 6, pp271-280.

[45]

[46]

[47]

[48]

[49]

[50]

[51]

[52]

[53]

[54]

[55]

120

Flanigan, G. A., "Oil Spill 'Fingerprinting’ with Gas
Chromatography," Research/Development, Sept, 197 7, pp. 28-36.

Mann, D. 0., "Comparison of Automotive Gasolines Using Capillary
Gas Chromatography I: Comparison Methodology," Journal of
Forensic Sciences, Vol. 34, July, 1986, pp.606-615.

Sanders, W. N., and Maynard, J. B., "Capillary Gas Chromatographic
Method for Determine the Hydrocarbons in Full-Range Motor
Gasolines," Anal. Chem. Vol. 40, No. 3, March 1968, pp.527-535.

Mach, M. H., "Gas Chromatography-Mass Spectrometry of Simulated
Arson Residue Using Gasoline as an Accelerant," Journal of Forensic
Sciences, Vol. 17, 1976, pp. 348-357.

Stone I. C., Lomonte, J. N., "Accelerant Detection in Fire Residues,"
Journal of Forensic Sciences, Vol. 18, 1977, pp. 78-83.

Clark, H. A., and Jurs, P. 0., "Classification of Crude Oil Gas
Chromatogram by Pattern Recognition Techniques," Anal. Chem., Vol.
51, No.6, May, 1979, pp. 616-623.

Loscalzo, P. J ., and DeForest, P. R., "A Study to Determine the Limit
of Detectibility of Gasoline Vapor From Simulated Arson Residues."
Journal of Forensic Sciences, Vol. 22, June 1979, pp. 162-167.

Mann, D. 0., "Comparison of Automotive Gasolines Using Capillary
Gas Chromatography II: Limitations of Automotive Gasoline
Comparison In Casework," Journal of Forensic Sciences, Vol. 32, July
1986, pp.616-628.

Saferstein R. and Park S. A., "Application of Dynamic Headspace
Analysis to Laboratory and Field Arson Investigations." Journal of
Forensic Sciences, Vol. 27, 1982, pp. 484-494.

Russell L. W., "The Concentration and Analysis of Volatile
Hydrocarbons in Fire Debris Using Tenax GC," Journal of Forensic
Science Society, Vol.21, 1981, pp.317-320.

Twibell J. D., "A Comparison of the Relative Sensitivities of the
Adsorption wire and Other Methods for the Selection of Accelerant
Residues In Fire Debris," Journal of Forensic Science Society, Vol. 22,
1982, pp. 155-160.

[56]

[57]

[58]

[59]

[60]

[61]

[62]
[63]

[64]

[65]

[66]

[67]

[68]

[69]

121

Tontarski, R. E., and Strobel R. A., "Automated Sampling and
Computer Assisted Identification of hydrocarbon Accelerant," Journal
of Forensic Sciences, Vol. 27, 1982, pp. 710-714. _

Andrasko, J ., "The Collection and Detection of Accelerant Vapors
Using Parous Polymers and Curie Point Pyrolysis Wires." Journal of
Forensic Sciences, Vol. 28, 1983, pp. 330-334.

Bentz, A. P., Analytical Chemistry, Vol. 48, No. 6, May, 1976, p. 454.

Gruenfeld, M. and Frank, U., "Prevention Behavior, Control,
Cleanup," Oil Spill Conference, May, 1977.

Siegel, J. A., "Solving Crimes with 3-D Fluorescence Spectroscopy,"
Anal. Chem., Vol. 57, 1985, p.934A.

Kubic, T. A., and Dwyer, J ., "Individualization of Automobile Engine
Oils I: The Introduction of Variable Separation Synchronous
Excitation Fluorescence to Engine Oil Analysis," Journal of Forensic
Sciences Vol. 28, No.1, Jan. 1983, pp186-199.

Holand, J. F., Anal. Chem., Vol. 55, 1983, p. 997A.
Grimths, P. J ., Anal. Chem., Vol. 55, 1983, p. 1361A.

Samuelsson, R, DeHaseth, J. A., and Azzaraga, L. V., Anal. Chem.,
Vol. 52, 1980, p.2215.

Chrestie, W. H., Anal. Chem., Vol. 53, 1980, p.13.

Kubic, T. A., "Individualization of Automobile Engine Oils II:
Application of Variable Separation synchronous Excitation
Fluorescence to the Analysis of Used Automobile Engine Oils,"
Journal of Forensic Sciences, Vol. 28, No. 2, April 1983, pp. 345-350.

"Oil Spill Identification by Fluorescence Spectroscopy", Oil Spill
Identification System, US. Coast Guard Research Center, Groton,
CT, June 1977, pp E1-E13.

Thrusto, A. D., "Characterization of Crude and Residual Oils by F8,"
Environ. Sci. Technol., Vol. 5, Jan. 1971, pp. 64-69.

Smith, H. F., "Identification of Crude Oils by SES," Anal. Chem., Vol.
48, No.3, March 1976, pp.520-524.

[70]

[71]

[72]

[73]

[74]

[75]

[7 6]

[77]

122

Rossi, T. M., and Warner, I. M., "Pattern Recognition of Two-
Dimensional Fluorescence Data Using Cross-Correlation Analysis,"
Applied Spectroscopy, Vol. 39, No.6, 1985, pp.949-959.

Files, L. A., "Gasoline and Crude Oil Fingerprinting Using Constant
Energy Synchronous Luminescence Spectrometry," Microchemical
Journal, Vol. 35, 1987, pp.305-314.

Siegel, J. A., "Fluorescence of Petroleum Products 1. Three
Dimensional Fluorescence Plots of Motor Oils and Lubricants,"
Journal of Forensic Sciences, Vol. 30, No.3, July 1985, pp. 741-759.

Siegel, J. A., "Fluorescence of Petroleum Products II. Three
Dimensional Fluorescence Plots of Gasolines," Journal of Forensic
Sciences, Vol. 32, No. 1, pp. 72-86.

Siegel, J. A., "Fluorescence of Petroleum Products III. Three
Dimensional Fluorescence Plots of Petroleum based Products,"
Journal of Forensic Seines, Vol. 33, No.6, Nov. 1988, pp. 1405-1414.

Wittkower, R. S., "Identification of Accelerant in Fire Residues by
Capillary Column GC," Journal of Forensic Sciences, Vol. 23, March
1978, pp.662-671.

Lucas, D. M., Journal of Forensic Sciences, Vol. 5, No.2, April 1960,
pp. 236-247.

Cadman, W., Journal of Forensic Sciences, Vol. 5, No. 3, July 1960,
pp. 369-385.

"[1? [it [[0040 "I" [1' [s