Вы находитесь на странице: 1из 30

DSC

Differential Scanning
C l i
Calorimetry
as used by AstraZeneca in Process
Safety Evaluations
Darren Maude
P
Process
Safety
S f t Group,
G
Pharmaceutical
Ph
ti l Development
D
l
t

Overview
1. Introduction
2. Pan/Crucible Selection
3. Instrument Calibration
4. Test procedure
5. What is a detected onset
6. Determination of detected onset
7. Endotherms preceding Exotherms
8. Effect
ff
off Ramp rate
9. Other Limitations
10 A
10.
Application
li i
off DSC D
Data to plant-scale
l
l operations
i
11. DSC Key Conclusions
2 | DSC | 03 November 2011

Introduction
DSC is widely used throughout the chemical industry as a
screening
g tool for thermal stability
y
AZ Process Safety Group have a total of 4 instruments
- DSC 821e

Main benefits of DSC for safety work seen as:


- Fast test turn-around
turn around ((~ 90 minutes)
- Small sample size
- Quantification of decomposition energy

However the limitations of DSC must be considered when


conducting safety assessments

The scope and limitations of DSC are the focus of this


p
presentation

3 | DSC | 03 November 2011

Pan/ Crucible Selection


High-pressure DSC pans MUST be used for process safety work
AZ use gold-plated
gold plated pans for maximum sample compatibility

Standard aluminum pans or other low


low-pressure
pressure pans must not be
used:
Leakage and endothermic vaporisation of volatile materials will
conceal exothermic activity

4 | DSC | 03 November 2011

Impact of Pan Selection


Tetrahedron Letters 48 (2007) 1233

H
N

Aluminium pan: H ~125 J/g


High
g Pressure Pan: H ~2600 J/g
g

N
N
H
N

N
N

5 | DSC | 03 November 2011

Aluminium pan: H ~600 J/g


High Pressure Pan: H ~1600 J/g

Instrument Calibration
Instrument needs to be correctly calibrated to accurately quantify energy of
decomposition
If the energy recorded is too low then potential explosivity of material may not
be identified later slide.
- T
Temperature
t
measurementt also
l requires
i
calibration
lib ti b
butt significant
i ifi
t errors ttend
d tto
occur less often.

Recommended that calibration is carried out with pan type to be used i.e. high
pressure gold plated at AZ.
-

Found that recommended method of universally calibrating with aluminium crucibles


showed a discrepancy (low energy value) when checked with the high pressure
crucibles used by AZ

As AZ typically only scan at 5K/min a simplified calibration is performed that is


only
y applicable at this rate. This avoids the need for a tau lag
g calibration ((i.e.
less parameters so a simpler and more accurate calibration achieved)

6 | DSC | 03 November 2011

Instrument Calibration

Within AZ, Macclesfield an indium and zinc sample are used to quantify
enthalpyy and temperature and results entered into the calibration file. Testing
g
water and benzoic acid in addition to re-running indium and zinc validates the
calibration (expect 5%)

Standards can deteriorate quickly on repetitive scanning and should be


replaced regularly. To prolong sample life:
Limit time standards are at elevated temperature i.e. limit scan range and
ensure standard removed from furnace asap after scan complete
Seal
S l standards
t d d under
d iinertt atmosphere
t
h
tto reduce
d
oxidation
id ti
Wrap standard in aluminium foil (indium and zinc will alloy with gold)

At AZ typically calibrate every 6 months and always after repair / maintenance

7 | DSC | 03 November 2011

Test Procedure
(R
(Ramped
d Temperature
T
t
Tests):
T t )
AZ generally use a temperature ramping rate of 5K per minute:
Ramp rate can affect detected onsets
- As shown in following slides.
Scanning tests should generally be run up to at least 400C.
Necessary to detect high energy decompositions occurring at high
temperatures
- Impurities can sometimes catalyse such events to occur at
lower temperatures
Will alert to potential (thermally stable) explosives
- AZ assume that
th t if decomposition
d
iti energy is
i <800J/g
<800J/ th
then th
the compound
d
will not have explosive properties (exception is azide compound)

8 | DSC | 03 November 2011

What is a Detected Onset?


Is:
The lowest temperature at which the instrument detects
exothermic
th
i activity.
ti it Thi
This iis d
dependant
d t on:
- Scan rate
- Sample size
- Pan selection
- Judgement of operator
- Slight
g variation between DSC instruments
Is not:
Where a reaction starts!
- Exothermic activity will still occur below the detected onset
temperature, but is not detected due to the relatively low
y of DSC.
sensitivity
- Onset temperatures detected by other instruments and
sustained self-heating on plant may occur well below the
detected DSC onset.
9 | DSC | 03 November 2011

Determination of Detected Onset


Onset must not be defined as used by crystallisation scientists
for melting point determination:
Taking tangents from steepest slope of exotherm and baseline
give values that are too high
high.
Mettler Software for Onset as part of the Integration function is
g
method:
based on tangents
- Do not use this method for safety studies
- Instead take left limit of integral
i.e. initial deviation from baseline

10 | DSC | 03 November 2011

Determining a DSC Onset


^ e xo

2/98 2/2

2 5 . 1 0 . 2 0 07 1 3 : 4 5 : 1 6

Me th o d : HP g o ld c r u c ib le ( 2 5 - 4 0 0 o C)
2 5 .0 - 4 0 0 .0 5 .0 0 /m in

20
mW

C o rre c t In te g ra l
W ro n g O n s e t"
t
"W
In te g ra l
1 2 .1 3 e + 0 3 m J
n o rm a lize d 1 0 3 2 .4 0 J g ^-1
Ons et
1 9 3 .1 2

O n s e t fo r p roc e s s
s a fe ty is le ft-lim it
o f e xo th e rm p e a k
O n s e t 1 6 1 .0 9

20
0

40

60
5

80
10

100
15

120
20

140

160
25

P S G C h a r n wo o d : M E TTL E R

11 | DSC | 03 November 2011

180
30

200
35

220
40

240

260
45

280
50

300
55

320
60

340

360
65

S TA R

380
70

m in

S W 8.10

Endotherms Preceding Exotherms


In some materials, an endothermic event directly leads into
an exothermic event
For Arrhenius decompositions, the endothermic event
may mask heat output at lower temperatures for the
exothermic event
event.
In quasi-autocatalytic decompositions, decomposition
leads to melting of the material after which much more
rapid decomposition begins.
- Self-heating that could be significant at the large scale can
occur below
b l
th melting
the
lti point:
i t
Dont assume that a sample is stable below its melting
endothem at any scale.
Prudent to take the melting endotherm onset as the
decomposition onset.
12 | DSC | 03 November 2011

Effect of Ramp Rate


AZ O
Oxime
i
Intermediate
I t
di t att 10C/min
10C/ i
^ e xo

2/9 45

0 3 . 0 2 . 2 0 04 0 9 : 0 9 : 2 7

As tra Ze n e c a O xim e In te rm e d ia te
H e a t ra m p rate 1 0 / m in u te

In te g r a l
6 4 7 5 .4 2 m J
n o r m a lize d 7 2 0 .2 9 Jg ^- 1
20
mW

E n d o th e rm
O n s e t 1 7 7 .1 1

E xo th e rm
O n s e t 1 8 9 .3 2
40
0

60

80
4

100
6

120
10

140
12

P S G C h a r n w o o d : M E TTL E R
13 | DSC | 03 November 2011

160
14

180
16

200
18

220
20

240
22

260
24

280
26

300
28

320
30

340
32

360
34

380

3 6 min

S TA R e S W 8 . 1 0

Effect of Ramp Rate


AZ O
Oxime
i
Intermediate
I t
di t att 1C/min
1C/ i
^ e xo

2/941-2

2 5 . 1 0 . 2 0 07 1 3 : 5 1 : 2 9

As traZenec a O xim e Interm ediate


Heat ram p rate 1 / m inute

In te g ra l
8 6 8 3 .6 8 m J
n o rm a lize d 7 0 4 .2 7 J g ^ -1
5
mW

O ns et 151.15

40
0

20

60

80

40

60

100
80

120
100

1 40
120

P S G C h a r n w o o d : M E TTL E R
14 | DSC | 03 November 2011

160
1 40

18 0
160

200
180

220
200

240
220

260
240

28 0
260

3 00
2 80

320
3 00

3 40
320

360
340

380

min

S TA R e S W 8 . 1 0

DSC - Other Limitations


Small sample size
- DSC generally unsuitable for two
two-phase
phase mixtures
- Difficulties in test cell preparation in avoiding evaporation of volatile
solvents
- Within AZ, DSC is generally only used for thermal screening of
isolated intermediates and products
But this obviously requires other test equipment to safely evaluate other inprocess streams..

Does not detect gas generation


- Example of this limitation given on next slide

15 | DSC | 03 November 2011

AZ BOC-Intermediate
Mixture free from hazards for processing at 40-50 C
C?
^e xo

2/995-2

2 9 . 1 0 . 2 0 07 0 8 : 4 5 : 3 9

Astra Ze n e ca "BO C " In te rm e d ia te

O n s e t 1 5 1 .6 9

Integral
5663.14 mJ
normaliz ed 378.30 Jg^-1
Peak
274.02

20
mW

40
0

60
5

80
10

10 0
15

120
20

P S G C h a r n w o o d : M E TTL E R
16 | DSC | 03 November 2011

140

160
25

180
30

200
35

220
40

240

260
45

280
50

300
55

S TA R

320
60

min

SW 8.10

AZ BOC-Intermediate
Thermall Screening
Th
S
i
Unit
U it (TSu)
(TS ) test
t t shows
h
that
th t endotherm
d th
is
i
associated with the evolution of a large quantity of gas
AZ "BOC" Intermediate
Thermal Screening Unit Test
300

75
70
Oven Temp (C)

65

Sample Temp (C)


( C)

250

60

T baseline (C)
Pressure (bara)

200

50
45
40

150
35
30
25

100

20
15
50
10
5
0
0

10

20

30

40

50

60

70

Time (minutes)

17 | DSC | 03 November 2011

80

90

100

110

0
120

Pressure (ba
ara)

Temperature ((C)
T

55

AZ BOC-Intermediate
I li ti
Implications
for
f Drying
D i
and
d Storage
St
Raw material was to be supplied at 100s tonne per year quantities
and stored/ stockpiled for many months
Unstable material- decomposes without heat output but with
gas evolution (CO2 and isobutene)
This is not predicted by the DSC test where it was initially assumed that
the endotherm was simply
p y associated with material melt.

Further tests showed decomposition was in fact autocatalytic


i.e. Material stability decreases with time.
Drying times were shortened and the material was kept refrigerated.
Lined cardboard drums were used for storage to prevent a dangerous
pressure build-up.

18 | DSC | 03 November 2011

Application of DSC data to plant-scale


operations
ti

DSC is not an overly-sensitive technique


AZ assume a conservative
ti detection
d t ti limit
li it off ttypically
i ll 20 W/k
W/kg using
i a
scan rate of 5K/min and viewing on a scale of -2 to 2mW/g
These conditions give an acceptable compromise with respect to test time
and sensitivity.

Thermal runaway at the plant scale can occur with initial heat rate of
far less the DSC detection limit

Necessary to extrapolate the DSC result to determine a safe


operating temperature at the plant scale

A widely used rule of thumb is the 100K rule for non viscous
liquids
- i.e.
i Tsafe = TonsetDSC 100 C

But note, the 100K rule can occasionally fail!

19 | DSC | 03 November 2011

Application of DSC data to plant-scale


operations
ti

A more scientific approach would be to make use of the Semenov model In


AZ we use the Temperature of No Return (TNR) concept Conservative for DSC

Conservative for typical


10,000L vessel
U > 5W/m2/K
Affects cooling

Exponential heat from


chemical reaction

50-200kJ/mol
depending on direction
of extrapolation

Linear newtonian
cooling

20 | DSC | 03 November 2011

Application of DSC data to plant-scale


operations
ti

At an onset of 147C under the given


parameters the curves jjust intersect
p
mixture would stop self heating
TNR = 44C would need to subtract
103K reasonably close to 100K rule
However at onset temperatures below
this the lines will never intersect so self
selfheating could occur.

21 | DSC | 03 November 2011

Application of DSC data to plant-scale


operations
ti
At an onset of 135C the
curves dont intersect
Mixture could self heat from any
y
temperature
100K rule predicts ok at T<35C?
Onset limit to when 100K rule can be
applied
pp

Similar effect would occur if the


ambient temperature was
changed
h
d ffrom 25C tto say
35C (India) cooling reduces
separating the lines.

At AZ in this situation we would


consider more sensitive testing
or a Time to Maximum Rate
(TMR) approach.

22 | DSC | 03 November 2011

Application of DSC data to plant-scale


operations
ti
The opposite
pp
applies
pp
if the heat losses
of the vessel were larger
Example here demonstrates how the
increased heat losses of an AZ 100 litre
Large Scale Laboratory vessel
(measured at 0.05W/kg/K) shifts the
cooling line. This has the obvious effect
of increasing
g the temperature
p
of no
return. Very useful if natural heat losses
of vessels are known and allows data to
be specifically tailored to the particular
p
plant/vessel

23 | DSC | 03 November 2011

Application of DSC data to plant-scale


operations
ti

The TNR model assumes standard kinetics (zero order)

Semenov approach assumes no temperature gradients. i.e.


can only be applied to non viscous liquids. At AZ we use the
Th
Thomas
model
d l (heat
(h t loss)
l
) for
f solids
lid and
d viscous
i
liliquids
id
(conservative calculation indicates that applying100K rule
only allows processing of material at ~10cm in depth).

Many decomposition reactions tend to have some element of


autocatalysis which may need consideration (next slides)

Given large extrapolations with DSC relatively minor


autocatalysis should be covered by the conservative
activation energy used in these extrapolations

More prominent autocatalysis may need further work as


desc bed
described

24 | DSC | 03 November 2011

Autocatalysis
^e
exo
mW

16364 098a

1 5 .0 9 .2 0 0 6 1 2 :3 1 :2 6

S a m p le : A ZD 62 4 4 F re e S id e C h a in To s y lat e S a lt , 7 . 1 0 0 0 m g

DS C

100
90
80

A u to c a ta lys is

70
60
50

I n te g ra l
n o rm a liz e d
O nse t
P eak
Le f t Lim it
R ig h t Lim it

40
30

6789.68 m J
9 5 6 . 2 9 J g ^ -1
1 8 6 .2 6
1 9 1 .1 0
1 6 5 .2 1
2 1 9 .7 2

20
10
0
-10

D M aude

-20
-0
-30
0

20
2

40
6

60
10

12

80
14

16

100
18

20

P r o c e s s H a z a r d s S e c tio n : M E T T L E R

120
22

24

140
26

28

160
30

32

180
34

36

2 0 t h A p ri l 2 0 0 6
200

38

40

240

220
42

44

46

48 min

ST AR e S W 9 .0 0

Asymmetrical shape to peak showing sharp rise and slower decay is


indicative of autocatalytic behaviour and may prompt further testing.

25 | DSC | 03 November 2011

Autocatalysis Isothermal Tests

nth order plot illustrates typical decay due to concentration effect

Autocatalysis plot - the most simple explanation is that the product of the reaction actually
catalyses the reaction, the more product is produced the faster the reaction so it actually
accelerates with time. Therefore the rate is not only temperature dependant but also time (age /
thermal history of the substance).

Simple mechanism might be:


A
B (slow)
A+B
2B (fast)
- Ratio of the two determines the extent of autocatalysis.

26 | DSC | 03 November 2011

Autocatalysis

Peak power output can be measured and used in TNR


model using conservative activation energy
OR

Several isothermal experiments can be performed at


different temperatures. Peak power outputs can be plotted
against
g
reciprocal
p
absolute temperature
p
to determine an
actual activation energy. This can be used to make the TNR
model less conservative however care needs to be taken
for large extrapolations using non conservative parameters.

27 | DSC | 03 November 2011

DSC - Key Conclusions


Pan selection is critical:- always use high pressure pans for safety
work
Onset
O
t mustt be
b defined
d fi d as l
left
ft lilimit
it off exotherm
th
Corresponds to initial lift from baseline.

The onset quoted by the instruments integration function


based on a tangents method is irrelevant for safety.
Critical runaway temperature on plant will normally be below the DSC
detected onset; (scale factor required).
If an endotherm is closely linked to an exotherm then take the
endotherm onset as the decomposition onset
Decomposition can still occur below the materials melting point.

Consider evaluation of results with respect to the actual process


- Is there autocatalysis
28 | DSC | 03 November 2011

DSC - Key Conclusions


- Has a conservative activation energy been used
- How long is the extrapolation
- Taking everything into consideration is the result / the temperature deemed
to be safe actually conservative?

And finally- remember DSC is only a thermal screening tool


Not best suited to two-phase
two phase mixtures
mixtures.
Cannot detect gas generation

29 | DSC | 03 November 2011

Questions?

30 | DSC | 03 November 2011

Вам также может понравиться