Sie sind auf Seite 1von 194

Final Report for Project:

Long-term Prediction of Emissivity of Structural materials for


High Temperature Reactor Systems

Project Number: NEUP DE-NE0000743


Duration: January 14, 2014 – December 31, 2017
Total Funding Level: $799,117

Principal Investigator: Robert V. Tompson Jr.


University of Missouri
Nuclear Science and Engineering Institute
E2433 Lafferre Hall
Columbia, MO 65211
tompsonr@missouri.edu
(573) 882-2881

Collaborators:
Tushar K. Ghosh, University of Missouri, (co-PI)
Hangjin Jo, University of Wisconsin-Madison
Sudarshan K. Loyalka, University of Missouri, (co-PI)
Kumar Sridharan, University of Wisconsin-Madison (UWM-PI)
Dabir S. Viswanath, University of Missouri, (co-PI)

Students:
Faten N. Al-Zubaidi, University of Missouri
Jonathan L. King, University of Wisconsin-Madison
Shawn E. Nelson, University of Missouri
Kyle L Walton, University of Missouri
Table of Contents

Summary ..........................................................................................................................................3

Project Scope and Description .........................................................................................................4

Objective 1 .......................................................................................................................................6

Overview ..............................................................................................................................6

Procurement .........................................................................................................................6

Preparation ...........................................................................................................................7

Objective 2 .....................................................................................................................................11

Overview ............................................................................................................................11

Molten Salt (FLiBe) Corrosion ..........................................................................................11

Helium and Air Corrosion .................................................................................................15

Objective 3: Measurements of Hemispherical Emissivity .............................................................19

Overview ............................................................................................................................19

Experimental Apparatus.....................................................................................................19

Uncertainty Analysis ..........................................................................................................24

Calorimetric Emissometer Maintenance and Upgrades .....................................................31

Results: Short-term oxidization studies .............................................................................36

Results: Long-term oxidization studies .............................................................................38

University of Wisconsin-Madison Final Report ..........................................................................pg#

2
Summary:

The purpose of this research was to extend the recent short-term studies of emissivity of reactor materials

performed independently at the University of Missouri and University of Wisconsin to tests aimed at longer-

term predictions of emissivity. The University of Missouri measured hemispherical emissivity, which

provides an integrated measurement, while the University of Wisconsin measured spectral emissivity (2 to

10 µm wavelength) from which the integrated quantity can be obtained. Hemispherical emissivity is needed

for MELCOR, RELAP and FLUENT type heat transfer code calculations, while spectral emissivity is

needed for a detailed Monte Carlo Transport theory computation of heat transfer and fundamental modeling

verification. The objectives of the study were:

Objective 1: Defining materials for long-term emissivity in cross-cutting reactor applications -


procurement, preparation, and analyses of materials.
Objective 2: High temperature materials testing in various environments and post-test surface analysis
Objective 3: Independent hemispherical and spectral emissivity testing.
Objective 4: Development of models to gain a predictive capability for long-term emissivity and
application and validation in current RCCS research programs.

The research has been performed at two universities over a four-year term with the participation of students,

scientific staff, and faculty. The report describes progress in realizing the objectives, and the

accomplishments at the two universities. There was considerable progress with respect to Objectives 1-3.

The model development was redirected towards modeling of interrelationships between the spectra and

hemispherical emissivities and uncertainty analysis of total hemispherical emissivity measurements.

3
Project Scope and Description:

The Reactor Pressure Vessel (RPV) and internal reactor components rely partially on radiation from their

outer surfaces for cooling. In high temperature reactors or in the event of an unexpected high temperature

excursion, the dependence on radiation for the expulsion of heat from the system becomes all the more

important because of the fourth power temperature dependence of radiated heat according to the Stefan-

Boltzmann equation [1]. The key material parameter that dictates the extent of radiated heat is emissivity,

which is defined as the ratio of emissive power of the materials’ surface to that of an ideal black body.

Emissivity is a surface phenomenon (typically involving only a few microns of depth) and is dictated by

the materials’ surface chemical composition as well as the physical nature of the surface such as roughness,

porosity, and texture. Since oxidation and/or some form of surface corrosion will inevitably occur at high

temperatures during long exposures, it is critically important that these surface effects be taken into account

in long-term modeling of emissivity.

The purpose of this research was to extend the recent short-term studies of emissivity of reactor materials

performed independently at the University of Missouri and University of Wisconsin to tests aimed at longer-

term predictions of emissivity. The University of Missouri measured hemispherical emissivity, which

provides an integrated measurement, while the University of Wisconsin measured spectral emissivity (2 to

10 µm wavelength) from which the integrated quantity can be obtained. Hemispherical emissivity is needed

for MELCOR, RELAP and FLUENT type heat transfer code calculations, while spectral emissivity is

needed for a detailed Monte Carlo Transport theory computation of heat transfer and fundamental modeling

verification.

The scope of the proposed research has been to obtain total hemispherical and spectral emissivity data

for:

[1] LWR, HTGR, FHR RPV Applications: High temperature air exposure emissivities for A533B,
A508, Fe-9Cr-1Mo ferritic steels, and 316 stainless steel (outer surface of FHR containment vessel).

4
[2] FHR Core Internals Applications: Hastelloy-N, 316 stainless steel, and graphite after exposure to
exposure to molten fluoride salt, FLiBe (LiF-BeF2).
[3] SFR Core Internals Applications: 316 stainless steel after exposure to liquid sodium (for SFR in case
of loss of coolant).
[4] HTGR Core Internals Applications: Alloys 800H and IN617 after exposure to reactor grade helium.

And to address:

[5] Modeling and Long-Term Predictions: Incorporate data obtained in goals (1) through (4) into
predictive models for hemispherical and spectral emissivity for different materials and environments,
and use the results from the two methods of emissivity measurements synergistically.
[6] Application of Models to current RCCS Research Programs: Apply emissivity data and models
from the proposed research to ongoing programs on air cooled RCCS as a validation of the codes and
determine effects of changes in measured emissivity values in these facilities.

The logical path to accomplishing these objectives has been:

Objective 1: Defining materials for long-term emissivity in cross-cutting reactor applications -


procurement, preparation, and analyses of materials.
Objective 2: High temperature materials testing in various environments and post-test surface analysis
Objective 3: Independent hemispherical and spectral emissivity testing.
Objective 4: Development of models to gain a predictive capability for long-term emissivity and
application and validation in current RCCS research programs.

The research has been performed at two universities over a four-year term with the participation of students,

scientific staff, and faculty. The report describes progress in realizing the objectives, and the

accomplishments at the two universities. There was considerable progress with respect to Objectives 1-3.

The model development was redirected towards modeling of interrelationships between the spectra and

hemispherical emissivities and uncertainty analysis of total hemispherical emissivity measurements.

5
Objective 1: Defining Materials for Long-Term Emissivity in cross-
cutting reactor application – Procurement, Preparation, and Analysis
of Materials.

Overview

Candidate materials for select Generation IV reactors and light-water reactors were chosen

for emissivity effects for long-term operating environments. Both structural and reactor pressure

vessel materials were selected, but the emissivity of RPV materials of high-temperature gas

reactors (HTGR) are critical to the function of the RCCS. A few of these materials are common

among different reactor technologies. Many of the selected materials had benn already studied for

emissivity of air-ingress by the collaborating universities. Therefore, some of the materials for

long-term emissivity measurements were already available. For total hemispherical emissivity

measurements per the ASTM 835-06, specimens have specific size and geometry requirements.

Cutting specimens for total hemispherical emissivity measurements depends on whether the metal

is sheet or plate stock.

Procurement

The materials for the cross-cutting reactor applications consist of readily available alloys

except that the SA 508 and SA 533B. SS 316L and A387 Grade 91 were already in stock from

previous studies on total hemispherical measurements1; 2 Rolled sheets of Alloy 617, Alloy 800H,

and HR-120 were purchased from separate suppliers for the current study. In addition, sheets of

SS 304 and Nickel 200 were obtained for evaluation of the calorimetric emissometers by

comparing to previously measured results. Though HR-120 was not in the original proposal, it is

similar class of alloy to Alloy 800H provided by Haynes International. The treatments for Alloy

800H and Alloy 617 for Helium corrosion for VHTR applications were straight-forward enough
6
that HR-120 was considered for measurements if time permitted. For rolled-sheet metals, the

thicknesses from 0.010” to 0.020” are the easiest to heat with DC currents, but some of the more

exotic alloys are not available in these thicknesses. Table 1 summarizes the acquired materials for

this study with their nominal thicknesses. The low-alloy steels, SA 508 and SA 533B, were harder

to find. No supplier was found to carry either metal. Eventually, Idaho National Lab (INL) was

contacted and was able to provide a bar of low-alloy steel that was dual certified for both SA 508

and SA 533B. Low-alloy pressure vessel steels are not manufactured in rolled sheets, but rather in

plate/bar stock.

Preparation

Two different sets of samples were made for the total hemispherical emissivity

measurements and the analysis of the corrosion layer from treatment simulating long-term reactor

environments. For sheet-stock, abrasive water jet cutting can easily cut specimens. Metal coupons

measuring 0.5” long by 0.5” were water jet cut from sheets of Alloy 800H, Alloy 617, HR-120,

and SS 316L for scanning electron microscopy (SEM) and X-ray photoelectron spectroscopy

(XPS) analyses. Emissivity measurements were split up between two similar ASTM style

calorimetric emissometers. The calorimeter used in the studies of Maynard et al.3; 4, Gordon et al.5,

Keller et al.6, and Hunnewell et al.2 can measure strips up to 10” in length. Strips of SS 316L were

water jet cut into strips with dimensions 10.000” long and 0.500” wide. The new second

calorimeter can measure up to 12” long strips with no modification. Strips of Alloy 800H, Alloy

617, and HR-120 where water jet cut with the dimensions of 12.000” long and 0.500” wide. All

dimensions from water jet cutting conformed to a tolerance of ±0.007”.

7
Table 1: Summary for materials used in total hemispherical measurements.

Physical Nominal
Alloy Availability
Form Thickness
Alloy 800H Sheet 0.032” Purchased
Alloy 617 (Batch 2) Sheet 0.020” Purchased
HR-120 Sheet 0.063” Purchased
SS 304 Sheet 0.030” Purchased
SS 316L Sheet 0.025” In stock
Nickel 200 Sheet 0.019” Purchased
SA 387 Grade 91 Batch 2 Plate ~0.510” In stock
A508/A533B Plate 0.75” Received from INL

Wire Electric Discharge Machining (EDM) was identified as the best way for cutting strips

from plate/bar stock. Capabilities can vary from shop to shop, thus a few EDM shops were

approached. One EDM shop was sent A387 Grade 91 to cut a test strip to compare with strips cut

from another shop. This EDM shop was chosen to cut 10.000” × 0.500” × 0.035” with a tolerance

of ±0.005” for each dimension. Excellent results were obtained with the thickness and width being

out of parallel up to 0.002”. This was mainly due to having the strips cut a thickness of 0.035”

instead of 0.015”–0.020” for A387 Grade 91 used in earlier studies. Due to the cost of EDM

machining, it was decided not to prepare A387 Grade 91 as its emissivity has already been

measured previously.

Additionally, a commercial lapping and polishing company was used to prepare a few

strips of A508/A533B with a mirror finish. The polishing and lapping was done by Valley Design

Crop. Initially, it was decided to have a matte finish on the A508/A533B to simulate an as-received

8
surface and eliminate surface effects from EDM machining. Since most of the samples were going

to be oxidized, it was decided to only have three strips of A508/A533B to save on cost for having

all strips lapped and polished. The A508/A533B was lapped to a thickness of 0.025'' ± 0.002''. The

strips were polished to a mirror like finish.

Samples that were water jet cut were carefully deburred and cleaned prior to corrosion

treatments. Edges of the strips for emissivity measurements were deburred with smooth cut file to

remove most of the burr. A diamond file with a grit size of 170-200 was then used to slowly remove

the remainder for the burrs. A508/A533B strips did not have any burrs to remove. An ultra-sonic

bath was acquired for this project to clean metallic strips used in the emissivity measurements.

Figure 1 shows the ultra-sonic bath used in this study. A custom cart was built to make ultra-sonic

a portable self-contained system. The strips of Alloy 617, Alloy 800H, and HR-120 were cleaned

in the ultra-sonic bath in a 0.05% Liqunox and DI water solution for 20 mins. Strips were rinsed

with DI water and Acetone and allowed to air dry before being stored individually in clean

polyethylene square tubes. The square tube and plastic endcaps were washed in a 1% Liqunox and

DI solution. Tubes and caps were rinsed thoroughly with DI water and allowed to air dry.

9
Figure 1: Ultra-sonic bath for cleaning
metallic strips for emissivity
measurements.

10
Objective 2: High Temperature Materials Testing in Various
Environments and Post-Test Surface Analysis.

Overview

For this work, it is important to produce surfaces representative of the appropriate reactor

application for thermal emissivity measurements. The high-temperature testing matrix for

emissivity measurements is given in Table 2. For total hemispherical measurements, corrosion of

the test strips in liquid Na and molten salts proved difficult. The long-term corrosion of candidate

materials in air and He were addressed, but there were institutional delays in the procurement of a

three-zone furnace. At the time of this report, air oxidations of SS 316L in air have been completed

and samples are in the process of being analyzed for SEM and XPS. Air oxidation for

A508/A533B is in progress. Parameters for He corrosion have been determined and additional

work will contiure. The experimental matrix for the spectral emissivity measurements is given in

the corresponding final report by UW-Madison.

Molten Salt (FLiBe) Corrosion

Initially, molten salt corrosion treatments were pursued over liquid Na corrosion due to the

amount of materials purposed to be corroded. The facilities and equipment had to be developed at

MU for the handling of FLiBe molten salts and enable long term corrosion with the FLiBe molten

salt. The use of Be containing salts at the MU facilities required several University departments to

work together. Such coordination was difficult and led to the abandonment of molten salt corrosion

work. The efforts and approaches for molten salt corrosion for total hemispherical emissivity

samples is described below.

11
Table 2: Summary for materials used in total hemispherical measurements.

Corrosive Media Temperature, Duration Materials

SS 316L
Air 573 K, 3,000 hr A508/A533B
A387-Gr91

Alloy 617
He 973 K, 1,000 hr Alloy 800H
HR-120

Na
773 K, 1,000 hr SS 316L
(liquid)

SS 316L
Molten Salt
973 K, 1,000 hr Hastelloy N
(FLiBe)
Graphite

A static corrosion scheme was planned where specimens are placed in a sealed capsule of

FLiBe molten salt. This capsule would be placed in an atmospheric furnace with an inert gas to

limit the corrosion of capsule exterior. One design involved a reusable U-shaped capsule. The U-

shaped capsule allow the molten to be hydrostatically moved from on leg to another to either add

or extract samples. Simple drawing of the U-tube capsule is shown in Figure 2 (left). An isolation

valve separated the legs once the FLiBe was moved from one side to the other. This design requires

a valve to withstand being heated in an atmospheric oven. Heating tape on was considered but it

wouldn’t have the uniformity of a furnace.

12
Figure 2: U-tube corrosion
capsule for molten salts.

Due the toxicity and reactivity of the molten salts, a custom glovebox was considered. The

glovebox was intended for the preparation of the molten salts by melting LiF and BeF2 together in

the proper portions and placing and sealing the FLiBe salt and samples in metal capsules. This

would require a means of heat removal and TIG welding feedthroughs for the glovebox. Though

FLiBe molten salts have a melting temperature of 733 K, LiF temperature is 1118 K7. A single-

zone large bore tube furnace was considered to melt the molten salt for preparation and for long-

term a corrosion. An Inconel tube can be connected to a custom port on the glovebox. Figure 3

shows how the three-zone furnace would be connected for melting LiF and BeF2. Additional

cooling capacity for the glovebox would not be needed if heat transfer from the tube furnace to the

glove box is minimized. It was hoped this setup would almost eliminate the exposure and

contaminate for fluoride salts as they never leave the glovebox.

The design of the Inconel tube is given in Figure 4. The tube has two regions called the

heating zone and buffer zone. The buffer zone is the length of the tube to help cool the tube to a

13
Figure 3: A tube furnace connect to the
glovebox would allow the high temperature
heating for the melting of LiF and BeF2.

safe temperature at the glovebox connection. The end could have been actively cooled by the force

convection of air or water cooling if natural convection was insufficient. Insulating plugs also act

to limit heat transfer to the tube ends. This idea was borrowed from the three-zone furnaces use

for air/helium corrosion. The furnace would be disconnected from the glove box for long-term

corrosion runs.

Figure 4: The Inconel tube shows the is setup to reduce heat transfer to the tube
ends with insulating plug and a buffer zone.

14
Helium and Air corrosion

In previous studies on total hemispherical emissivity of reactor materials, short-term

oxidations were done inside the calorimetric emissometer. Long-term oxidations are not practical

in the calorimetric emissometer. Also using the calorimetric emissometer resulted in fouling of the

vacuum chamber and high-emissive coating on the bell jar interior. Since no controls for the

concentrations of impurities or main oxidizers was planned for gaseous corrosions treatments, a

simple setup using three-zone tube furnace with a controlled flow of helium/air provided by a mass

flow controller was built. The overall setup of the long-term gaseous corrosion is given in Figure

5. The system features a two-stage regulator with a two pigtailed manifold to enable the switching

of empty gas cylinders for long runs. Another important feature for the corrosion experiments was

the ability to corrode a batch of samples at a given time. Some samples were added and removed

to obtain oxidation for shorter times, while some samples remained for the full 1,000 hrs to 3,000

hrs. The oxidation times were so long that corrosion during heating and cooling for adding and

removing were neglected.

Figure 5: Custom sample boat for air or helium corrosion for a batch of metal strips
for emissivity measurements and coupons.

15
Fabrication and purchasing of the custom sample boat provided a few challenges. Initially,

a dense ceramic like alumina was chosen for the sample boat. The sample boat would have eight

bins to hold individually hold metal coupons for post-analysis testing. The walls of the ends were

raised above the body of the sample boat with eight slots to hold up to eight metal strips for total

hemispherical emissivity measurements. Several companies could not fabricate such a piece from

a dense ceramic. Injection molding was a possibility but turned out to be cost prohibitive.

Eventually, ground quartz was identified as an affordable material for the sample as temperatures

were under 1273 K. Due to the price, two sample boats were purchased with one as a backup, yet

the second sample boat was used for the additional furnace for air corrosion. The custom sample

boat for batch corrosion of metal coupons and strips are shown in Figure 6.

Figure 6: Custom sample boat for air or helium corrosion for a batch of metal strips
for emissivity measurements and coupons.

The procurement of a three-zone tube furnace was important for uniform heating of metals

strips up to 45.7 cm (18”) for air/he corrosion experiments. A three-zone providing ± 5 K

temperature uniformity over a length of 60.0 cm was purchased. A two-stage CGA 580 regulator

with a two-pigtailed manifold was purchased and a mass-flow controller with a He flow-rate of 0

to 73 L/min was purchased as well. The three-zone tube furnace for gaseous corrosion was as-

received, but damaged in shipment. The shipping claim and repairs of the furnace delayed

16
procurement of the furnace for a few months. Though a work order for upgrading the electrical for

the furnace was placed at the time of purchase, installation of the furnace was delayed even after

receiving it after repairs. The furnace was eventually installed. All metal lines were installed from

the regulator to the inlet of the furnace tube to eliminate the permeation of H2O and CO2 from the

environment during long term runs. Leak test with a mild soapy solution for leak testing found no

leaks. For helium corrosion, industrial grade helium was used to mimic the impure helium

environment of a HTGR. Due the abandonment of the molten salt corrosion and time left in the

project, a second three-zone furnace was acquired to do air and helium corrosions separately.

There were no issues in the acquisition and installation of the second furnace. It is nearly identical

for the first furnace but with a needle valve with a digital mass-flow meter to set flow rates. Figures

7 and 8 show the actual furnace setups used for He and air corrosion.

The air corrosion for SS 316L has been completed while the corrosion of air A508/A533B

is in progress during the writing of this report. SS 316L has been oxidized for times from 500 hr

to 3,000 hrs in steps of 500 hrs. Initially, three strips were oxidized for the same duration for

backup measurements whereas only two strips were oxidized for the same times as experience was

gained. There was little corrosion of the SS 316L and no detectable mass gain. The resulting oxide

was not stable being heated under vacuum when heated for emissivity measurements. The coupons

at the corresponding times were sent for XPS analysis for general composition and chemistry of

major constituents. Since XPS is surface sensitive, it is hoped information can be determined for

the thin oxide. A508/A533B was also oxidized for 500 hr and 1000 hr at 573 K. It was expected

for the A508/A533B to corrode more than the SS 316L, but the mass change is undetectable. The

temperature for air oxidation has been increased to 773 K for the remainder of the A508/A533B

samples to get sufficient corrosion. It was expected the low alloy steel to have more corrosion.

17
Figure 7: The first three-zone tube furnace being used for Helium corrosion.

Figure 8: The second three-zone tube furnace for long-term air oxidation.

18
Objective 3: Measurements of hemispherical emissivity

Overview

A brief description of the experimental apparatus and method are given here, together with

improvements, and data obtained, and results of uncertainty analysis. Both of the calorimetric

emissometers at MU have undergone maintenance over the course of this study. The older

calorimetric emissometer was cleaned after a long series of short term studies. An uncertainty

analysis of the calorimetric measurements showed the measurement of the DC current as a large

source of uncertainty in the emissivity. More precise DC transducers were obtained for both

systems to reduce the uncertainty in emissivity measurements. The spectral directional emissivity

measurements are reported UWM’s final report.

Experimental Apparatus

Two calorimetric emissometers were built during the short-term studies. These systems

were in accordance to the ASTM standard ASTM C835-06. The newer calorimetric emissometer

was initially built to corrode samples for calorimetric emissivity measurements. It was later

converted to a calorimetric emissometer with a slight modification in an attempt measure apparent

emissivity in low pressure environments. The systems are similar in design with differences in ball

jar size, DC current rating and power capacity for sample heating, and data acquisition hardware.

The basic design of the calorimetric emissometers are show in Figure 9. A metal strip (3)

is clamped between two electrodes inside a stainless-steel bell jar (1a) and base well (1b). Macor®

rods (2) and metal tabs were used as an interconnect between the spot-welded thermocouples (TC)

on the specimen (0.07 mm OD) to heavier gauge TC wire (0.15 mm OD) to the vacuum

19
Figure 9: An illustration of the main elements of
the emissivity measurement setup and electrical
circuit for specimen heating: (1a) vacuum chamber
base well, (1b) vacuum chamber bell jar (high
emissivity coating on inner wall); (2) ceramic
thermocouple mounting posts for easy connection
of specimen thermocouples to thermocouple
feedthroughs; (3) sample strip; (4) dead weight to
prevent sample strip from buckling; (5) current
shunt resistor; (6) DC power supply; (7) polarity
reversing double-pole double-throw switch; (8)
hall effect sensor, (9) mechanical roughing pump;
and (10) turbo-molecular pump. Note
thermocouple feedthroughs, system vent, pressure
gauges, and data acquisition are not shown.

20
feedthroughs. The interior of the bell jar was coated with a highly-emissive paint—Aremco HiE-

CoatTM 840-M. The bell jar can be raised and lowered to load samples. A turbomolecular pump

(11) with a rotary-vane backing pump can achieved pressures of 1.3 × 10−4 Pa to 8.0 × 10−5 Pa.

The bottom clamp moved freely to accommodate thermal expansion of specimen. A deadweight

(6) was used to keep the sample under tension. A ceramic threaded coupler (5) prevented heating

of the deadweight. The actual electrical connection of the bottom clamp to the bottom support was

from a low-impedance flexible cable. A programable linear DC source (8) provided the DC current

for Joule heating of specimens. Earlier studies by the group used a shunt resistor (7) in series with

the test specimen for measuring DC currents. A hall-effect sensor (8) was added to measure DC

currents with less noise. The hall-effect sensor was upgraded to a fluxgate magnetometer for

greater precision. Joule heating with DC currents introduces an offset in the specimen TC’s

electromotive force (EMF) creating a bias in temperature readings. A polarity-reversing switch

(9) is used to reverse the direction of the DC current. Averaging specimen temperature readings in

both DC current directions eliminates this offset giving the actual temperature.

Total hemispherical emissivity is determined when temperatures reach a steady-state after

changing the applied DC current through the test specimen. To simplify calculation of total

hemispherical emissivity, ASTM C835-06 has the following condition for the relation of specimen

surface area (𝐴1 ) and emissivity (𝜀1 ) to chamber interior surface area (𝐴2 ) and emissivity (𝜀2 ):

1 𝐴1 1
≫ ( −1), Eq. 1
𝜀1 𝐴2 𝜀2

where 𝜀2 > 0.8 and 𝐴1 ⁄𝐴2 < 0.01. The interior surface of the bell jar and base well for the older

and newer calorimetric emissometers 11,850 cm2 and 13,885 cm2 respectively. Using the

minimum emissivity for the HiE-CoatTM 840-M of 0.90 and total surface area of a typical strip of

21
65.8 cm2, the right-hand side of Eq. (1) yields 5.6 × 10−3 and 4.7 × 10−3 for the older and newer

calorimetric emissometer. The emissivity of the test section can be obtained by the Stefan

Boltzmann formula:

𝑄1
𝜀1 = , Eq. 2
𝜎2𝐿1 (𝑤1 + 𝑡1 )(𝑇14 − 𝑇24 )

where 𝑄1 is the heat radiated by the test section in W; 𝐿1 , 𝑤1, and 𝑡1 are the length, width, and

thickness of test section in m; 𝑇1 is the average temperature of test section in K; 𝑇2 is the average

temperature of chamber in K; and 𝜎 is the Stefan-Boltzmann constant W m−2 K−4 (Ref. 15).

Figure 10 shows the section within the five TC pairs over the center of the test specimen.

The emissivity can be determined at any of the voltage drops or from the addition of voltage drops

between adjacent TCs. The voltage drops were measured by adding voltage taps to the alumenl

leads of the TC. The area between TC #2 and TC #4 was the preferred section for emissivity

calculations and consistent with the definition of the test section in ASTM C835-06.

Figure 10: Total hemispherical emissivity of Alloy 718 oxidized in air at for up to
10 min.

22
The heat radiated along a test section with determined from the following version of Ohm’s Law:

𝑄1 = 𝐼|∆𝑉24 | , Eq. 3

where 𝐼 is the applied DC current in A and ∆𝑉24 is the voltage drop between TC #2 and TC #4.

This is due to offsets present in the signals. These offsets are different in nature for the respective

signals, but are both eliminated through averaging.

In addition to statistical averaging of data, some quantities required averaging to either

remove offsets in the signals due to the direction of the DC current and spatial averaging. The

direction depended readings will yield the true reading. As described earlier, specimen

temperatures get an offset from the electric field applied to the strip. The true temperature was

obtained from averaging the temperature in the forward and reverse directions:

𝑇𝑗,A + 𝑇𝑗,B
𝑇𝑗 = Eq. 4
2

where 𝑗 is the index of specimen TC (𝑗 = 1, … 5), A is the index of the DC current in the positive

direction, and B is the index of the DC current in the negative direction. For the hall effect sensor

and fluxgate magnetometer, background magnetic fields and thermal effects also cause offset in

the output. While a measurement with no applied DC current can determine the value of this

offset, averaging this signal in the A and B directions:

𝑉HES,A + 𝑉HES,B
𝑉HES = Eq. 5
2

Temperature along the test section and in the chamber were sampled at several locations. The

temperatures at the different locations were spatially averaged as follows:

23
𝑁
𝑇𝑖
〈𝑇〉 = ∑ Eq. 6
𝑁
𝑖=1

where 𝑇𝑖 is the temperature in the 𝑗 th location in °C and 𝑁 is the number of TCs. Two sources of

uncertainty arise in Eq. 6 due to the propagation of uncertainty from each 𝑇𝑖 and the standard

deviation of each 𝑇𝑖 with the spatial average 〈𝑇〉.

Uncertainty Analysis

A rigorous methodology of determining uncertainty analysis in the total hemispherical

emissivity measurements has been one of the achievements in this study. ASTM C835-06

examines the uncertainty in the Stefan-Boltamann equation used for the calculation of total

hemispherical emissivity from the input variables. It also relies on a limited intralaboratory

comparison of a reference materials to establish uncertainty and precision of the method. This was

the approach used by Maynard and Maynard et al. (Ref. 3 and Ref. 4). Recently, Fu et al.8 reported

uncertainties in their total hemispherical emissivity data obtained in a vacuum calorimeter similar

to that described in ASTM C835-06. Yet, no details are given for calculations of such uncertainties.

In addition, Fu et al.8; 9 used AC currents for their Joule heating, which does not cause the artifacts

in the specimen temperatures associated with DC currents. It thus appeared necessary to determine

uncertainties from the measured raw data to the calculated emissivity. Details of the uncertainty

analysis will be reported separately in a journal publication. The results are described briefly here.

The determination of uncertainty followed JCGM 208:100, or The Guide to the Expression

of Uncertainty in Measurements, or GUM. The GUM method propagates uncertainty in the

function 𝑓(𝑥) using the expectation value and variance of its Taylor’s expansion. Consider the

quantity 𝑦 calculated from correlated input quantities (𝑥1 , 𝑥2 , … 𝑥𝑁 ) by the function:

24
𝑦 = 𝑓(𝑥1 , 𝑥2 , … , 𝑥𝑁 ) . Eq. 7

Ignoring higher order terms in the Taylor series of 𝑓(𝑥1 , 𝑥2 , … , 𝑥𝑁 ), the uncertainty in 𝑦 is

obtained from Law of Propagation of Uncertainty (LPU):

𝑁 𝑁−1 𝑁

𝑢𝑐2 (𝑦) = ∑ 𝑐𝑖2 𝑢2 (𝑥𝑖 ) + 2 ∑ ∑ 𝑐𝑖 𝑐𝑗 𝑢(𝑥𝑖 )𝑢(𝑥𝑗 )𝑟(𝑥𝑖 , 𝑥𝑗 ), Eq. 8


𝑖=0 𝑖=1 𝑗=𝑖+1

where 𝑐𝑖 and 𝑐𝑗 are sensitivity coefficients of 𝑓(𝑥1 , 𝑥2 , … , 𝑥𝑁 ) due to the ith and jth variables

respectively, 𝑢 are standard uncertainties, and 𝑟 is the correlation coefficient between 𝑥𝑖 and 𝑥𝑗 .10

Sensitivity coefficients are the partial derivatives of 𝑓(𝑥1 , 𝑥2 , … , 𝑥𝑁 ) with respect to the ith or jth

variable. They represent the change in 𝑓(𝑥1 , 𝑥2 , … , 𝑥𝑁 ) due to the change in 𝑥𝑖 and weigh each

uncertainty component’s contribution to the combined uncertainty. The second term in Eq. 8 is

zero for independent input quantities. In the GUM approach, the functional relationship between

the measured quantity and input quantities is called the measurement model. The measurement

model for calorimetric measurements consists of Eq. 1 and Eqs. 3 to 6. Measured quantities whose

magnitude was not affected by the direction of applied DC current were combined into a single set

for analysis.

Since the GUM applies the LPU to only standard uncertainties, an important aspect of the

uncertainty analysis requires converting stated errors specifications into standard uncertainties. An

assumption of the GUM method is that all input quantities are random variables. Error limits are

assumed to be a type of confidence interval or coverage interval of a probability distribution. The

JCGM 208:100 provides some conservative guidelines in assigning probability distributions and

estimating confidence levels of error limit. Another resource is the NASA handbook NASA-

HDBK-8739.19-3 (Ref. 11). The sources of uncertainties and their values are compiled into what

25
is called an uncertainty budget. Only in simple measurements where uncertainties have a single

value can all the information for the error sources and uncertainties can be tabulated together in a

single uncertainty budget. Uncertainties tend to vary with the value of the measured quantity. Due

to the number of variables and intermediate calculations in the calorimetric measurements, a

simple budget cannot be given for the entire measurement.

An example of this process is given for the uncertainty in the temperature measurements.

Sources of error in the thermocouples themselves and the temperature DAQ reading the TC. The

USB-TC temperature DAQ has a maximum error stated of 0.691 °C.12 The manual describes this

value combined from errors in linearization, cold-junction compensation, and system noise. With

no other information, the error values are considered 95% confidence intervals of a normal

distribution. Another source is in the resolution for the analog to digital converter (ADC) of the

USB-TC. Resolution errors are considered rectangularly distributioned as the least significant digit

in the reading has an equal probability of being rounded up or down from the following digit.11

Rectangular distributions have 100% confidence level as the lower and upper values contain all

values. Manual specifies USB-TC resolution the in mV, this is converted to degrees Celsius using

the inverse polynomials for Type K thermocouples.13 Uncertainty in the Type K TCs should ideally

be obtained from calibration to a reference or measurement of its Seebeck coefficient.14 Without

such information, the standard tolerance in for Type K TCs is used for TC error:

max(±2.2 °C, 0.0075 𝑇), Eq. 9

where 𝑇 is the nominal value of temperature in °C.13 As a tolerance, the values from Eq. 9 are the

error limits of a rectangular distribution. The last source of error is in the repeated temperature

measurements. The standard uncertainty is obtained from statistical sampling of data. Table 3

26
summarizes the sources in error. The divisors in Table 3 convert the confidence intervals of their

respective distributions to standard deviations.

Table 3: Sources of error in temperature measurements.


Error Probability
Divisor
Source Limits Probability Distribution

K-Type TC Eq. 9 100% Rectangular √3


USB-TC Accuracy 0.691 °C 95% Normal 1.96

USB- TC Resolution 1.2 × 10−4 °C 100% Rectangular √3

Repeatability Student-t

Again, temperatures measurements will be used to demonstrate the calculations for

combined and propagated uncertainties. The combined uncertainty in a temperature measurement,

from the sources in Table 3 was:

2 2 2
𝑢𝑇 ( 𝑇 ) = √𝑢K–type ( 𝑇 ) + 𝑢USB–TC,ACC + 𝑢USB–TC,RES + 𝑢2𝑇,RAN , Eq. 10

where 𝑇 is the temperature in °C, 𝑢K–type is standard uncertainty from Type K TC tolerance in

°C,. 𝑢USB–TC,ACC is the standard uncertainty from USB-TC accuracy in °C, 𝑢USB–TC,RES is the

standard uncertainty from the USB-TC resolution in °C, and 𝑢𝑇,RAN is the standard uncertainty

from repeated measurements in °C. For specimen TCs, the uncertainty from offset averaging was

obtained by applying the LPU to Eq. 4:

𝑢2𝑇 (𝑇𝑗,A ) + 𝑢2𝑇 (𝑇𝑗,B )


𝑢𝑇,𝑗 = √ . Eq. 11
4

27
where 𝑇𝑗,A and 𝑇𝑗,B are the temperature readings in K for the positive and negative DC current

directions for the 𝑗 𝑡ℎ TC, respectively. Data for the chamber temperatures are combined into one

set for mean and standard deviation. Spatial averaging has two general uncertainty sources. The

first is the combined uncertainty that is propagated through spatial average (Eq. 6) and the second

is the deviation of the quantity at each measured point from the spatial average itself. The

uncertainty in the spatial average is:

𝑁𝑇 𝑁𝑇
2 𝑢2𝑇,𝑘 (𝑇𝑘 − 〈𝑇〉)2
𝑢〈𝑇〉 =∑ 2 +∑ , Eq. 12
𝑁𝑇 𝑁𝑇 (𝑁𝑇 − 1)
𝑘=1 𝑘=1

where 𝑁𝑇 is the number of locations, 𝑢𝑇,𝑘 is the standard uncertainty in 𝑘 𝑡ℎ temperature in °C, 𝑇𝑘

is the temperature at the 𝑘 𝑡ℎ location in °C, and 〈𝑇〉 is the spatial average from in °C.

The final uncertainty in the emissivity measurements is examined for as-received Alloy

718 Run #2 and Hastelloy X. Applying the LPU to Eq. 2, yields the combined uncertainty in total

hemispherical emissivity. The uncertainty contribution from the uncertainty in the Stefan-

Boltzmann constant is neglected. Table 4 shows the uncertainty budget for the emissivity of Alloy

718 at the lowest and highest temperatures. Heat generation was the dominant source of

uncertainty in the emissivity. From Table 4, the relative uncertainty in the emissivity goes from

13.26% to 2.15% from the lowest to highest temperatures. The relative uncertainty in the heat

generated is 13.52% at the lowest temperature to 2.08% at the highest temperature. Temperature

was the second main source of uncertainty. The uncertainty for specimen temperature was low

from the offset averaging. The increase of the uncertainty in chamber temperatures was from the

28
Table 4: Uncertainty budget for the emissivity calculation for Alloy 718 Run #2
between TC #2 and TC #4.

Pt Qty ̅
𝒙 𝒖𝒙̅ 𝒄𝒙̅,𝒊

𝐴1 6.659 × 10−4 m2 2.4651 × 10−6 m2 −271.777 m–2

𝑇1 563.63 K 0.62762 K −1.408 × 10−3 K–1

1 𝑇2 310.03 K 0.94211 K 2.3427 × 10−4 K–1

𝑄1 0.624 W 0.08437 W 0.28857 W–1

𝜺 0.181 0.024 —

𝐴1 6.659 × 10−4 m2 2.4651 × 10−6 m2 −411.419 m–2

𝑇1 1274.66 K 1.8609 K −8.659 × 10−4 K−1

2 𝑇2 368.08 K 4.04448 K 2.0851 × 10−5 K−1

𝑄1 27.12 W 0.56332 W 1.0102 × 10−2 W−1

𝜺 0.2740 0.0059 —

larger non-uniformity between the thermocouples on the chamber walls at higher temperatures.

This non-uniformity could be variation of the natural convection outside of the chamber. Figure

11 shows the emissivity and its standard uncertainty for Alloy 718 and Hastelloy X. The blue data

points are uncertainties in excess of 5% as required by the ASTM C835-06 standard. The apparent

change in uncertainty over the measured temperatures was due to overall electrical resistance

required to heat the test specimens.

The uncertainty in the heat generated was dominated by the uncertainty in electrical current

measurements. Table 5 shows the uncertainty budget of the heat generated for the highest

temperature measured for as-received Alloy 718 Run #2. Again, the relative uncertainty in the DC

current was 2.07%, which is like 2.08% relative uncertainty for the heat generated. Here the

29
Table 5: Uncertainty budget for the heat generated between TC #2 to TC #4 at the
highest temperature measured.

Qty Unit ̅
𝒙 𝒖𝒙̅ 𝒄𝒙̅,𝒊

𝛥𝑉24 V 4.1960 × 10−1 6.92445 × 10−4 64.6443 W V−1

𝐼 A 64.64 1.3383 4.1960 × 10−1 W A−1


𝑄1 W 27.12 0.56 —

Figure 11: The emissivity and standard uncertainty of Hastelloy X versus Alloy
718. The points with light blue are uncertainties greater than 5%. The Hastelloy X
requires less DC current to heat.

voltage drop added a slight uncertainty to the heat measurements. The hall-effect sensor was the

largest contributor to DC current measurements with a fixed uncertainty of 1.88 A, which was

substantial for smaller DC currents. Around 27 A, the uncertainty in the DC current approaches

~5%. In Figure 11, Alloy 718 required around 65 A to heat to maximum temperature, whereas

Hastelloy X only needed 34 A. The ASTM C835-06 also requires heat measurements to be within

1%. The data fails this requirement even at higher temperatures where the uncertainty satisfies the

30
overall uncertainty in the measurement. The analysis showed that more precise measurement in

DC currents was required to reduce the uncertainty in emissivity measurements overall. Fluxgate

manetgometers were acquired to improve uncertainty in DC current measurements.

Calorimeter Maintenance and Upgrades

The older of the two calorimetric emssiometers had minor maintenance until but poor

adhesion of black body coating on the bell jar. The application of HiE-840M coating to bell jar

was not thoroughly documented and several approaches were used to coat the bell jar—including

a buying a new bell jar. After some trial and error, a successful coating was obtained. Due to this

down time, it was decided to reservice or replace vacuum hardware and thoroughly clean the

vacuum system of the emissometer. A few equipment failures occurred after overhauling the

system, but backups were available to continue the work.

After the conclusion on-going measurements during the onset of this, the older calorimetric

emissometer was cleaned including reapplying of blackbody coating and rewiring of data

acquisition hardware. This coating failed overtime and issues were encountered in reapplying a

new coat. The process used to apply the 840-M to the bell jars for the previous works could not be

successfully reproduced. This included applying 840-M with a foam brush while the bell jar was

vertically suspended. This had little adhesion despite a thorough degreasing and preparing the

surface with an etchant made for the 840-M. Several glass viewports on the bell jar were found to

be damaged upon disassembly. It was decided to acquire a new bell-jar with only one viewport

with a stainless-steel blank. Second major attempt to paint the bell jar involved applying 840-M

with a small air-brush. Up to 4 thin coatings were applied with an overnight slow curing between

each coat. This coating only lasted for a few months after doing a few runs on Hastelloy X. The

31
third and final attempt involved sanding the bell jar with 60-grit SiC pads. The coating was sanded

with 100-grit SiC paper to remove loose paint and applied again. This took three tries to get

complete coverage without any further issues to date.

Figure 12: The older calorimetric emissometer after installation of new bell-jar.

During the painting of the bell jar, the vacuum system was overhauled. This included using

getting a new roughing gauge and sending the cold-cathode gauge for reservice. The base well and

vacuum connections were thoroughly cleaned to remove deposited films. The upper and lower

specimen holders were clear of oxide layers with glass shot. The turbomolecular pump was

reserviced as it’s service history was unknown and many years of use. Due to the new gauge, it

was determined that the system could be roughed pump in 45 minutes until 50 mtorr was reached

32
to turn on the turbopump. Approximately 8 × 10-7 torr was readily achievable when starting

measurements. Figure 12 shows the calorimetric emissometer after final coating for bell jar.

The newer calorimetric emissometer was used until midway in the study. Several issues

with the system developed since the system was last used. The DAQ card for both analog voltage

and temperature was not working and was replaced with a National Instruments chassis with

temperature and analog voltage cards. The chassis has extra slots for additional cards. The ceramic

posts for connecting specimen TCs was damaged from use. They were made from an alumina

foam/fiber boar, Zicar RS-99R. Due to its cost, it was decided to go back to Macor® for the ceramic

posts. The TCs for measuring gas temperatures in the chamber for apparent emissivity

measurements were removed. The new Macor ® rods were also adjusted for extra the TCs. A new

upper support rod for the top clamp/electrode allows the measurement of longer test specimens. A

new roughing gauge operable to atmospheric pressures was obtained to better monitor rough

pumping of the system. The cold cathode gauge was reading incorrectly, it was reserviced and

calibrated.

The new roughing gauge showed the current mechanical pump was inadequate to rough

pump the system. Rough pumping to ~50 mtorr for the turbopump took up to three hours.

Cleaning and greasing the bell jar O-ring and replacing mechanical pump oil did not have any

impact on rough pumping times. After several more tests, it was determined the pump rate, ~7

CFM, was inadequate for the bell jar. A new roughing pump with roughly twice the pumping rate

was purchased. With the new pump, it took approximately an hour to go under 70 mtorr. After

arranging vacuum feedthroughs for the revised setup, slight leak developed in the system affecting

the ultimate pressure down to 6 × 10-5 torr. The minimum required by the ASTM standard 9.75 ×

33
10-6 torr. A helium leak detector was used to isolate the leak but could not be pin pointed. Trail

and error of suspected leaks were the solved the problem.

Figure 13: Total hemispherical emissivity of Alloy 718 oxidized in air at for up to 10
min.

Over the course of the study a LabView program was developed to simplify operation of

the calorimetric emissometers and create a standard data format for calculations. Previous work

on the older calorimetric emissometer used two programs to manage data acquisition. Since the

temperature DAQ had LabView libraries, a single LabView program was written to read and

record data from the DAQ devices. The reduced the complexity of the data taking process

minimized user errors and training time. The program allows simple changes to DAQ the hardware

and sampling rate of data from temperature and voltage DAQs. Data from either DAQ can be

34
viewed, but temperatures are used to determine steady-state conditions. The program automatically

generates file names of each new measurement by adjusting a few simple controls. A common

issue is accidently writing over the most recent measurement. The program detects if there is no

new change in the file name and forces the user to update the file name parameters. The program

can be used on any similar system with only changes to the DAQ libraries (sub-VIs) for particular

DAQ hardware.

Figure 14 LabVIEW user interface for the calorimetric emissometers monitoring readings
and acquiring data.

Following the uncertainty analysis of emissivity measurements, high-precision fluxgate

magnetometers were obtained to replace the hall-effect sensors on both systems. Low-ripple DC

power units were built for the fluxgate magnetometers. The fluxgate magentometers output a small

electric current proportional to the measured current. Ten Ohm high-precision resistors were

chosen to measure secondary current of the fluxgate sensors directly with a DAQ card.

35
Results: Short-term oxidation studies

Several studies on candidate materials from the continuing work were completed during

onset of this project. This involved additional measurements of Alloy 718 to determine the cause

in the drop of emissivity in oxidized Alloy 718. Figure 15 shows the emissivity four different

samples of Alloy 718 oxidized in air for 10 minutes. Each run was terminated at different

temperatures around the peak emissivity at 1200 K. Examination of the samples with SEM-EDS

could not find any changes in surface morphology and composition to explain the behavior in

oxidized Alloy 718. Additional study on the emissivity of SS 347 was in progress when the long-

term oxidation studies started for short-term oxidation and roughness effects. Total hemispherical

emissivity was obtained only for as-received SS 347, which is in Figure 16. The total hemispherical

emissivity for as-received SS 347 varied from 0.34 to 0.42 over the temperature range 510 K to

1089 K.

Figure 15: Total hemispherical emissivity of Alloy 718 oxidized in air at for up to
10 min.

36
Figure 16: Total hemispherical emissivity of SS 347 as-received.

The short-term oxidation and related studies that have been completed during the present

project have resulted in the following publications:

 Keller, B. P., Nelson, S. E., Walton, K. L., Ghosh, T. K., Tompson, R. V., and
Loyalka, S. K., “Total hemispherical emissivity of Inconel 718,” Nuclear
Engineering and Design. 287, 11-18 (2015)

 Azmeh, C. B., Walton, K. L, Ghosh, T. K., Loyalka, S. K., Viswanath, D. S., and
Tompson, R. V., “Total Hemispherical Emissivity of Grade 91 ferritic alloy with
Various Surface Conditions,” Nuclear Technology. 195, 87-97 (2016)

 Hunnewell, T. S., Walton, K. L., Sharma, S., Ghosh, T. K., Tompson, R. V.,
Viswanath, D. S., and Loyalka, S. K., “Total Hemispherical Emissivity of SS 316L
with Simulated Very High Temperature Reactor Surface Conditions,” Nuclear
Technology. 198, 293-305 (2017).

 Walton, K. L., Maynard, R. K., Loyalka, S. K., Ghosh, T. K., Tompson, R. V.,
Viswanath, D. S., and Loyalka, S. K., “Total Hemispherical Emissivity of Potential
Structural Materials for Very High Temperature Reactor Systems: Alloy 617”.
(Submitted, under review)

37
Results: Long-term oxidation studies

Test runs using as-received Hastelloy X were done after the installation and coating of the

new bell jar for the calorimeter. Figure 17 shows the data two different strips of Hastelloy X from

the same sheet (0.010” thick). The data is compared with Hastelloy X data from Maynard et al.4

whose samples were from a different sheet of Hastelloy X with a thickness of 0.020”. The new

data shows that reproducible results were obtainable with two different operators. There are

differences compared to the older data. Some of this variation may be due to the older data being

obtained with a shunt resistor. This configuration resulted in noisy current measurements.2 Though

acceptable values of the current were obtained with averaging. Rejecting the outliers were

sometimes required, which can shift the average value. The Hall-effect sensor does not suffer

significant noise.

Figure 17: Total hemispherical emissivity of Alloy 718 oxidized in air at for up to
10 min.

38
Total hemispherical emissivity on the long-term corrosion in air has been obtained at the

writing of this report. Figure 18 shows the emissivity for as-received SS 316L from this study

compared to Hunnewell et al. 2017. The two data sets agree quite well indicating the emittance

calorimeter is in working condition after the major maintenance. The use of standard uncertainties

determined for this method allow better comparison for the two data sets.

Figure 18: Total hemispherical emissivity of as-received


SS316L data obtained in present paper compared with
Hunnewell et al.’s data for SS316L.

Despite the thin oxidized layers for long-term oxidation of oxidation of SS 316L at 573K,

the duration of oxidization has noticeable effect on the emissivity of SS 316L. The emissivity of

as-received SS 316L is compared to oxidation times 500 hr to 1500 hr and 2000 hr to 3000 hr in

Figures 21 and 22 respectively. As noted earlier, the oxide layers of corroded SS 316L were lost

during the measurement of emissivity. Though retested data is within 2% as permitted by the

standard over the measured temperature, it is believed oxide is lost where the data plateaus above

39
1050 K. This plateau is seen for all oxidized samples. The oxidation of SS 316L at 500 hrs and

1,000 hrs are quire similar with an overall increase in emissivity of 0.04 at 420 K to 0.14 at 1050

K. The emissivity drops for the oxidation at 1,500 hrs. The data at 1,500 hrs of emissivity of 0.01

at the lower temperatures and 0.02 at the higher temperatures measured.

Figure 19: Total hemispherical emissivity for SS316L


oxidized in air at 573 K for 1500 h.

40
Figure 20: Total hemispherical emissivity for SS316L as-
received & oxidized in air at 573 K for 500 h,1000h, and
1500h.

Figure 21: Total hemispherical emissivity for SS316L as-


received & oxidized in air at 573 K for 2000 h,2500h, and
3000h.

The following publication are presently under preparation:

 Al-Zubaidi, F. N., Walton, K. L., Ghosh, T. K., Tompson, R. V., and Loyalka, S.
K., “Long-term effects of air oxidation on Total Hemispherical Emissivity of SS-
316L” (Manuscript in preparation)

 Walton, K. L., Ghosh, T. K., Tompson, R. V., and Loyalka, S. K., “Uncertainty
Analysis of Steady-State Calorimetric Emissometers” (Manuscript in preparation)

41
References

1. AZMEH, C. B. et al., "Total hemispherical emissivity of Grade 91 ferritic alloy with various surface
conditions," Nuclear Technology, 195, 1, 87, (2016); https://doi.org/10.13182/NT15-54.
2. HUNNEWELL, T. S. et al., "Total Hemispherical Emissivity of SS 316L with Simulated Very High
Temperature Reactor Surface Conditions," Nuclear Technology, 198, 3, 293, (2017);
https://doi.org/10.1080/00295450.2017.1311120.
3. MAYNARD, R. K. et al., "Hemispherical Total Emissivity of Potential Structural Materials for Very
High Temperature Reactor Systems: Haynes 230," Nuclear Technology, 179, 3, 429, (2012);
https://doi.org/10.13182/NT11-5.
4. MAYNARD, R. K. et al., "Total Hemispherical Emissivity of Potential Structural Materials for Very
High Temperature Reactor Systems: Hastelloy X," Nuclear Technology, 172, 1, 88, (2010);
https://doi.org/10.13182/NT10-6.
5. GORDON, A. J. et al., "Hemispherical total emissivity of Hastelloy N with different surface conditions,"
Journal of Nuclear Materials, 426, 1-3, 85, (2012); https://doi.org/10.1016/j.jnucmat.2012.03.026.
6. KELLER, B. P. et al., "Total hemispherical emissivity of Inconel 718," Nuclear Engineering and Design,
287, 11, (2015); https://doi.org/10.1016/j.nucengdes.2015.02.018.
7. SOHAL, M. S. et al., "Engineering Database of Liquid Salt Thermophysical and Thermochemical
Properties," INL/EXT-10-18297, Idaho National Laboratory (2010);
https://doi.org/10.2172/980801.
8. FU, T., TAN, P., and PANG, C., "A steady-state measurement system for total hemispherical emissivity,"
Measurement Science and Technology, 23, 2, (2012); https://doi.org/10.1088/0957-
0233/23/2/025006.
9. FU, T., TAN, P., and ZHONG, M., "Experimental research on the influence of surface conditions on the
total hemispherical emissivity of iron-based alloys," Experimental Thermal and Fluid Science, 40,
Supplement C, (2012); https://doi.org/10.1016/j.expthermflusci.2012.03.001.
10. BIPM et al., "Evaluation of measurement data—guide for the expression of uncertainty in measurement.
JCGM 100: 2008," Citado en las, (2008); (current as of.
11. "Measurement Uncertainty Analysis Principles and Methods, NASA Measurement Quality Assurance
Handbook - ANNEX 3," NASA-HDBK-8739.19-3, (2010-07-13);
<https://standards.nasa.gov/standard/nasa/nasa-hdbk-873919-4> (current as of 28 Nov. 2017).
12. MEASUREMENT COMPUTING CORPORATION. "USB-TC: Specifications." Rev 10A edition,
2016. <https://www.mccdaq.com/pdfs/manuals/USB-TC.pdf> (current as of 22 Nov. 2017).22
Nov. 2017.
13. "Standard Specification and Temperature-Electromotive Force (emf) Tables for Standardized
Thermocouples," E230/E230M-12, ASTM International, West Conshohocken, Pa;
https://doi.org/10.1520/E0230_E0230M-12.
14. BENTLEY, R. E. Theory and Practice of Thermoelectric Thermometry. vol. 3, Springer, 1998.
Handbook of Temperature Measurement.

42
Final Report for Project:

Thermal Emissivity of Nuclear Reactor


Candidate Alloys
U.S. DoE - NEUP Contract No.: DE-NE0000743.

Project Investigators: Dr. Kumar Sridharan, Dr. Hangjin Jo,


Jonathan King, Kyle Blomstrand, Richard Bisson

University of Wisconsin, Madison

Project Contact: Kumar Sridharan (kumar@engr.wisc.edu)


ii
i

Abstract
To achieve higher efficiencies, future nuclear reactor concepts are aimed at
operating temperatures significantly higher than current light water reactors (LWR).
At these high operating temperatures or during accidental high temperature
excursions in LWR, the dependence on thermal radiation for the expulsion of heat
from the system becomes critically important because of the fourth power
temperature dependence of radiated heat according to the Stefan-Boltzmann
relationship. The material property that dictates the extent of radiant heat is
emissivity, which is defined as the ratio of radiant heat from a material’s surface to
that of a black body. While rough approximations of emissivity have sufficed for light
water reactor heat transfer analysis for normal operating conditions, a deeper
understanding is necessary for heat transfer predictions of Gen IV high temperature
reactors or in the case of high temperature excursions in LWR.

A thorough and flexible mastery in any physical problem is achieved only


when a strong predictive capability grounded in theory is established. To this end, it
is desirable to develop first-principles-based modeling for reactor alloy emissivities.
However, challenges remain. First, the necessary foundation of experimental
reference of reactor emissivity is still under development. Until this point, exposure
studies of emissivity have been limited to air-oxidation and no detailed study has
evaluated the emissivity of samples after exposures to molten salt, liquid sodium, or
helium gas despite the prevalence of these environments in Gen IV reactors.
Second, emissivity of nuclear reactor alloys presents a particularly challenging
modeling problem. The emissivity of an oxidizing alloy can vary by nearly an order of
magnitude through exposure. Here, dependencies of composition, temperature,
roughness, oxidation, emission angle, and wavelength must be simultaneously
considered. First principles have yet to be used to describe the overall emissivity of
a surface so complex. In this work, the current state of the art is advanced along
both the experimental and modeling frontiers.
ii
Steels and nickel-based alloys were exposed to environments of air, He gas,
molten FLiNaK salt, molten FLiBe salt, liquid sodium or supercritical CO2 before
being measured using at least one of four different radiometric methodologies.
Studies on SA508 (reactor pressure vessel material, RPV) and T91 (common
codified reactor structural material that is also being considered as RPV material for
HTGR) emissivity during air oxidation was conducted in collaboration with the
National Renewable Energy Laboratory using a SOC-100 Hemispherical Directional
Reflectometer. For the first time, spectral emissivity was collected for reactor alloys
exposed to environments of molten salt (SS316, Hastelloy-n), He gas (Haynes-230,
Inconel 617), and liquid sodium (SS316). For this latter work, methodologies for
reflective radiometric and emissive radiometric techniques were developed and
applied to the respective exposed candidate materials using a Bruker VERTEX 70
FTIR system.

A Geometric Optics Approximation was developed and applied to predict


emissivity of platinum, tungsten, and AISI 304 stainless steel surfaces while
accounting for composition, temperature, roughness, emission angle, wavelength,
and polarity. The model achieved good agreement with various experimental results
collected during the past decades. This research demonstrated how a complex
directional spectral emissivity model may be sampled to culminate in total emissivity
while maintaining various dependencies. The modeling approaches developed here
are relevant and transferrable to reactor alloys.
iii

Contents

Abstract ............................................................................................................................. i 
List of Tables ................................................................... Error! Bookmark not defined. 
1  Introduction............................................................................................................... 1 
1.1  Reactor Pressure Vessel Alloys .................................................................. 5 
1.2  Structural Materials ................................................................................... 10 
1.3  Experimental Techniques .......................................................................... 11 
1.4  Modeling Efforts ........................................................................................ 14

2  Experimental Methods ............................................................................................ 19 


2.1  Test Materials ............................................................................................ 19 
2.1.1  Low Alloy Steels ................................................................................. 19
2.1.1.1  High Temperature He Gas .................................................................. 27
2.1.1.2  Molten Salt .......................................................................................... 27
2.1.1.3  Liquid Sodium ..................................................................................... 29
2.1.1.4  Supercritical CO2................................................................................ 30
2.2  Measurement Apparatus ........................................................................... 31 
2.2.1  ET100 Emissometer (Reflective Radiometry) ..................................... 31 
2.2.2  SOC-100 HDR w/ Nicolet FTIR (Reflective Radiometry) .................... 33 
2.2.3  Bruker VERTEX 70 FTIR with Hyperion Microscope (Reflective
Radiometry) ........................................................................................................... 35 
2.2.4  Bruker VERTEX 70 FTIR (Emissive Radiometry) ............................... 36 
2.3  Sources of Error and Uncertainty .............................................................. 40 
2.3.1  Error in Bruker Reflectance ................................................................ 41 
2.3.2  Error in Bruker Emittance ................................................................... 42 
2.3.2.1  Validation of the Meausrement Approach ........................................... 42 
2.3.2.2  Measurement Uncertainty ................................................................... 48 
2.3.2.2.1  Variation in Surface Temperature ....................................................... 49 
2.3.2.2.2  Error in Reference Emissivity.............................................................. 52 
2.3.2.2.3  Error in System Response .................................................................. 54 
iv
2.3.2.2.4  Overall Uncertainty ............................................................................. 54

3  Results and Analysis – RPV Materials (Low-alloy Steels) ...................................... 56 


3.1  Spectral Directional Measurements of SA508 ........................................... 56 
3.1.1  Roughened SA508 measured at Room Temperature......................... 57 
3.1.2  Exposed SA508 measured at Room Temperature ............................. 58 
3.1.3  Unexposed SA508 measured at Elevated Temperatures ................... 59 
3.2  Spectral Directional Measurements of T91 ............................................... 61 
3.2.1  Roughened T91 measured at Room Temperature ............................. 61 
3.2.2  Exposed T91 measured at Room Temperature .................................. 62 
3.2.3  Unexposed T91 measured at Elevated Temperatures ....................... 63 
3.3  Discussion – Low Alloy Steels ................................................................... 64 
3.3.1  Estimation of SA508 Total Emissivity from Data ................................. 64 
3.3.2  Estimation of T-91 Total Emissivity from Data .................................... 70

4  Results and Analysis – Structural Alloys ................................................................ 73 


4.1  Helium Gas Environment .......................................................................... 73 
4.2  Molten Salt Environment ........................................................................... 75 
4.3  Liquid Sodium Environment .................................................................... 777 
4.4  sCO2 Environment .................................................................................... 78

5  Modeling ................................................................................................................. 81 


5.1  Theory ....................................................................................................... 81 
5.1.1  Evaluation of ε(s,λ,θ,T) at a surface point s ...................................... 833 
5.1.2  Evaluation of ε(λ,θ,T) across the surface ............................................ 84 
5.1.3  Evaluation of ε(λ,T) ............................................................................. 84 
5.1.4  Evaluation of ε(T) ................................................................................ 85 
5.2  Model Results ........................................................................................... 86 
5.2.1  Case 1: Flat Platinum Surface ............................................................ 89 
5.2.2  Case 2: Flat 304 SS Surface .............................................................. 92 
5.2.3  Model setup for roughened surfaces .................................................. 95
v
5.2.3.1  Pseudorandom Surface Profile Generation ........................................ 95 
5.2.3.2  High-angle Scattering ......................................................................... 95 
5.2.4  Case 3: Roughened Platinum Surface................................................ 98 
5.2.5  Case 4: Roughened SS 304 Surface ................................................ 101
5.2.6 Case 5: Oxidized/Roughened SA 508 surface ……………………
105

6  Conclusion.........................................................................................................11009 
6.1  Experimental I: RPV Materials .............................................................11009 
6.2  Experimental II: Structural Alloys .........................................................11211 
6.3  Theoretical Modeling ............................................................................11312

Appendix A – Comprehensive Sample Matrix ........................................................... 1143


Appendix B: Comparison of Radiometric Reflection Data sets ................................... 116 
Appendix C: Bruker Emittance vs Other Methods ....................................................... 123

References ................................................................................................................. 126 


vi

[THIS PAGE IS INTENTIONALLY LEFT BLANK]


vii

Nomenclature

Abbreviations
RPV Reactor Pressure Vessel
VHTR Very High Temperature Reactor
LWR Light Water Reactor
MSR Molten Salt Reactor
RCCS Reactor Cavity Cooling System
FTIR Fourier Transform Infrared Spectroscopy
GOA Geometric Optics Approximation
BR Bruker Reflectance
BE Bruker Emittance
RCWA Rigorous Coupled Wave Analysis
FDTD Finite Difference Time Domain
MEEP MIT Electromagnetic Equation Propagation
BRDF Bi-directional Reflectance Distribution Function
EDM Electrical Discharge Machining
ASME the American Society of Mechanical Engineers
SEM Scanning Electron Microscope
HDR Hemispherical Directional Reflectometer
NREL National Renewable Energy Laboratory
MSBR Molten Salt Breeder Reactor

Expressions Subscripts
q heat flux Q heat transfer s-polarization
λ wavelength x position (perpendicular)
T temperature L,l Length p-polarization
kb Boltzmann constant k thermal conductivity (parallel)
E blackbody spectral h heat transfer coefficient, λ spectral
radiance Planck constant
σ Stefan-Boltzmann P Radiative Power i incident
Constant
k imaginary refraction S Emission Signal N normal
index (Experimental) incidence
R reflectivity Ω Solid Angle R Reference,
Room Temperature
r reflection ratio F Temperature Variation m measured
Error Term
n real refractive index Q Overall Uncertainty e Environmental
(Experimental) Temperature
Ω solid angle λ Wavelength
θ zenith angle Re Reynold’s Number
φ azimuthal angle ε Emissivity or Emittance
A Area S Period shift (Modeling)
viii
σR RMS roughness Ψ Model roughness
factor parameter
p probability

[THIS PAGE IS LEFT INTENTIONALLY BLANK]


ix

List of Figures

Figure 1.1. Two Concentric cylinders. The inner cylinder has temperature T1, radius r1, and
emissivity ε1. The outer cylinder has temperature T2, radius r2, and emissivity ε2. The gas (tan color)
between the two surfaces exhibits a thermal conductivity of k

Figure 1.2. Spectral emissivity progression across time for an iron sample oxidized in air at 480oC.
Constructive and destructive interference by oxide is visible

Figure 1.3. Spectral emissivity progression of oxidizing SA508 at 500 °C (left) and 700 °C (right)

Figure 1.4. Planck Distribution of Emissive Power. As temperatures increase, Emissive power
increases across all wavelengths, with greater relative increases occurring at lower wavelengths

Figure 1.5. Cumulative Distribution Function of E(λ,T). At greater temperatures, the proportion of
energy emitted at shorter wavelengths increases

Figure 1.6. Heat transfer simulations performed by Hong et al. for a 270kW Helium heater scaled
experiment. The core in the radiation-neglected case reaches a temperature 500 °C greater than that
of the radiation-included simulation

Figure 2.1. Lindberg furnace used to oxidize samples in high temperature ambient air environments

Figure 2.2. SEM cross section image of SA508 oxidized at 350 °C over 10 hours. Red arrows indicate
the oxide layer. Detachment appears to have occurred along the right side of the surface

Figure 2.3. SEM cross section image of SA508 oxidized at 600 °C over 5 hours. Red arrow indicates
the oxide layer. Detachment appears to have occurred through most of the oxide layer

Figure 2.4. SEM cross section image of T91 oxidized at 750 °C over 5 hours. Red arrow indicates the
oxide layer

Figure 2.5. Inconel box furnace used for FLiNaK molten salt corrosion tests
x
Figure 2.6. Atmosphere-controlled glove box housing FLiBe corrosion test

Figure 2.7. Liquid sodium loop testing facility for corrosion testing of samples used in this study

Figure 2.8. Photograph of the SC-CO2 autoclave system used in this study

Figure 2.9. Surface Optics ET-100 Emissometer in carrying case

Figure 2.10. Surface Optics ET-100 Emissometer connected to desktop measurement software.
Samples are placed on top of the uncovered golden integrating sphere (currently covered with cap in
the image) and then measured

Figure 2.11. SOC-100 Hemispherical Directional Reflectometer Coupled to Nicolet FTIR, managed by
NREL Concentrating Solar Power Group

Figure 2.12. Bruker VERTEX 70 FTIR coupled with Hyperion microscope

Figure 2.13. Schematic of VERTEX 70 emission measurement methodology

Figure 2.14. Normalized Planck Distributions

Figure 2.15. Preliminary emissivity measurements of Haynes-230 reference sample superimposed w/


normalized Planck distribution functions. Noise develops when Planck grows weaker and also when
background competes (longer λ)

Figure 2.16. 120 grit – finished SA508, the Roughest Consensus Sample. Good agreement is
achieved between the three reflectance methods. Every sample measured by the three having a
smoother surface than the RCS also achieved good consensus while for samples that were rougher
than RCS, Bruker reflectance data diverged from the NREL FTIR data

Figure 2.17 Spectral normal emissivities for Inconel 617 after 500-hour exposure to 1000 °C helium
gas with oxidative impurities and reference Haynes 230 determined using the simple method and the
gold subtraction method

Figure 3.1. Emittance measurements of SA508 air-oxidized at 300°C and 600°C. Measurements were
taken at room temperature

Figure 3.2. Emittance measurements of roughened SA508

Figure 3.3. Emittance measurements of SA508 air-oxidized at 350°C

Figure 3.4. Reference SA508 measured at elevated temperatures. Tm* indicates the second 100 °C
measurement following high temperature characterization. The large increase between the two
measurements indicates oxidation occurred during measurement
xi
Figure 3.5. Emittance measurements of roughened T91

Figure 3.6. Emittance measurements of T91 air-oxidized at 750°C for 5 hours

Figure 3.7. Emittance measurements of T91 at elevated temperatures

Figure 3.8. Estimated total angular emittance coefficients of SA508. Values are calculated using
equations 1-4 on data from samples oxidized at 350°C for various exposure times

Figure 3.9. Calculated ε(θ=10,T=350°C) of SA508 samples through various points of air-exposure

Figure 3.10. Estimated total angular emittance coefficients of SA508. Values are calculated using
equations 3.1-3.4 on the data collected for the sample oxidized at 600°C for five hours. Experimental
data was acquired at room temperature

Figure 3.11. Estimated total normal emittance coefficients of SA508. Values are calculated using
equations 3.1-3.4 on the data collected by Cao for the sample oxidized at 500°C for four hours (a)
and for the sample oxidized at 700°C for five hours (b). Experimental data was acquired in-situ at
500°C and in-situ at 700°C, respectively

Figure 3.12. Estimated total angular emittance coefficients of T91. Values are calculated using
equations 1-4 on the data collected for each sample used. The exposed sample was subjected to
750°C air for 5 hours before measured at room temperature. The bare samples were unexposed
mirror-finished samples measured while at 300°C and 25°C, respectively

Figure 4.1. Spectral normal emissivity for Haynes 230 after 500-hour exposure to 1000 °C helium
gas. Measurements taken at 300 °C. Photos of the samples are provided to the right of the spectra

Figure 4.2. Spectral normal emissivity for Inconel 617 after 500-hour exposure to 1000 °C helium gas.
Measurements taken at 300 °C. Photos of the samples are provided to the right of the spectra

Figure 4.3. Spectral normal emissivity for 850°C FLiNaK – exposed Hastelloy-N. Measurements
taken at 300 °C. Refer to Table 2.4 for additional info

Figure 4.4. Spectral normal emissivity for 850°C FLiNaK – exposed SS-316. Measurements taken at
300 °C. Refer to Table 2.4 for sample info

Figure 4.5. Spectral directional emissivity for 750°C LiF-BeF2 – exposed SS-316 at 15 degree
incidence. Measurements taken at room temperature using Bruker Reflectance. Refer to Table 2.4 for
sample info

Figure 4.6. Spectral directional emissivity for 650°C Na – exposed SS-316. Measurements taken at
300 °C. Refer to Table 2.4 for sample info
xii
Figure 4.7. Spectral normal emissivity for 650°C supercritical CO2 – exposed Haynes 230.
Measurements taken at 300 °C. Refer to Table 2.4 for sample info

Figure 4.8. Spectral normal emissivity for 650°C supercritical CO2 – exposed stainless steel 316.
Measurements taken at 300 °C. Refer to Table 2.4 for sample info

Figure 5.1. The progression of the GOA towards a spectral emittance. The first step (1) involves
projecting a ray of light onto the input surface and evaluating for the reflected and absorbed intensity.
The Fresnel Equations at the particular spot on the profile are used for this evaluation, as explained in
section 2.1. The second step (2) involves starting the incident ray at a new location on the plane of
incidence and performing the same progression. The average of satisfactorily many evaluated
starting locations, uniformly distributed across the incident plane, yields the directional emittance for
that surface as explained in section 2.2. A new incident angle is evaluated in the third step (3). At this
new incident angle, a new spectral directional emittance is evaluated using steps (1) and (2). An
adequate number of directional emittances may be integrated to yield a total spectral emittance as
explained in section 2.3

Figure 5.2. Normal Spectral Emissivity Coefficients of Platinum. The model assumes temperature
independence. Good agreement is achieved between the model and the works of Silvera and Rolling.
Model values tend to be slightly less than experimental values

Figure 5.3. Parallel Component of Platinum Relative Directional Spectral Emissivity Coefficients
through various wavelengths. Rolling experimental (left) and model output (right) both show increases
at longer wavelengths, particularly at higher angles of incidence

Figure 5.4. Perpendicular Component of Platinum Relative Directional Spectral Emissivity


Coefficients. Rolling experimental (left) and model output (right) both show little to no change across
wavelength

Figure 5.5. Model and Experimental values [84][85][86][24][87][74]for total hemispherical emissivity of
platinum

Figure 5.6. Parallel Component of 304 Stainless Steel Relative Directional Spectral Emissivity
Coefficients. Rolling experimental (left) and model output (right) both show increases at longer
wavelengths, particularly at higher angles of incidence. T = 953 °K

Figure 5.7. Perpendicular Component of 304 Stainless Steel Relative Directional Spectral Emissivity
Coefficients. Rolling experimental (left) and model output (right) both show little to no change across
wavelength
xiii
Figure 5.8. Total hemispherical emissivity coefficients of SS 304 and Tungsten. Values are generated
through Monte Carlo- sampled GOA for a smooth surface and are compared to the output of
Rosenberg

Figure 5.9. Parallel Component of Emitted Light from (a) Rolling and (b) model output at wavelengths
of 1.5 μm and from (c) Rolling and (d) model output at wavelengths of 6 μm. Several roughness
scales were evaluated

Figure 5.10. Perpendicular Component of Emitted Light from (a) Rolling and (b) model output at
Wavelengths of 1.5 μm and from (c) Rolling and (d) model output (top) and at Wavelengths of 6 μm.
Various roughness scales were evaluated

Figure 5.11. Model and Experimental values [84][85][86][24][87][74] for total hemispherical emissivity
of platinum. Model inputs of Ψ = π/40 and γ=.02 were used

Figure 5.12. Parallel Component of Relative Spectral Directional Emittance of polished stainless steel
304 at 953 °K (left) and model output where T = 953 °K, Ψ = π/200, and γ = .01 (right). The
perpendicular component remained unchanged from case 2

Figure 5.13. Total Directional Emittance of 304 Stainless Steel. Experimental results from Rolling are
shown in (a) and (c) while model output is provided in (b) and (d)

Figure 5.14. SS 304 Total Hemispherical Emissivity Coefficients from Experiment[74][90]and Model
Output

Figure 5.15. (a) Tungsten Total Hemispherical Emissivity Coefficients from


Experimental[91][92][93][94][95][96] and Model Output and (b) Zoomed-in Tungsten Total
Hemispherical Emissivity Coefficients from Experimental[92][94][95] and Model Output

Figure 5.16. Measured spectral optical constants: (a) refractive index and (b) distinction coefficient

Figure 5.17. Comparison of experimentally measured spectral emissivity and predicted emissivity by
the computational model for (a) mirror polished reference and oxidized samples at 350 oC and (b)
oxidized samples at 300 oC and 600 oC

Figure 5.18. Comparison of experimentally measured spectral emissivity and predicted emissivity by
the computational model for roughened surfaces
xiv

List of Tables

Table 1.1. Scenario Evaluation for 600 MW(t) Prismatic Gas-Turbine Modular Helium Reactor

Table 2.1. Nominal Compositions of the Low Alloy Steels Tested

Table 2.2. Applications, Test Materials and Conditions

Table 2.3. Profilometry measurements of unoxidized samples

Table 2.4. Applications, Test Materials and Conditions

Table 2.5. Nominal Compositions of the Ni-based Alloys Tested

Table 2.6. Nominal Compositions of the Fe-based Alloy Tested

Table 2.7. Profilometry measurements of exposed samples

Table 2.8. ET-100 Incident Angles and Spectral Ranges

Table 2.9. Calculated sample surface temperatures [°C]

Table 2.10. ∆Ttotal for each Sample

Table 5.1. Details of Presented Model Evaluations for each if the four Cases

Table 5.2. Normal Spectral Emittance Coefficients 


xv

[THIS PAGE IS INTENTIONALLY LEFT BLANK]


1

1 Introduction 
There exist three canonical mechanisms of heat transfer: conduction, convection
and radiation. For a given object, heat loss occurs through these mechanisms.
Uncoupled, the transfer of energy associated with each mechanism can be
expressed through simple 1D equations.

The net heat flux associated with conduction, or heat transfer through the physical
contact between two objects, may be expressed as

  (1.1)

Where,
q = heat flux [W · m-2]
T = temperature [K]
x = position [m]
k = conductivity [W · m-1 ·K-1] (within dx)

The net heat flux associated with convection, or the transfer of energy between an
object and it’s environment due to fluid motion, may be expressed as

(1.2)
Where,
h = heat transfer coefficient [W · m-2]
Ta = temperature of object’s surface [K]
Tb = temperature of surrounding media [K]

The gross heat flux (or radiant power per area) associated with the emission of
electromagnetic radiation may be expressed through the Stefan-Boltzmann
relationship (Equation 1.3),
2

(1.3)
Where
ε = emissivity of the object [-],
σ = Stefan-Boltzmann Constant = 5.670367E-8 [W · m-2 · K-4]

With the net radiant heat flux between two surfaces (assuming parallel plate
geometry) behaving as

(1.4)

Where subscript 1 denotes the first surface and subscript 2 denotes the second
surface.

Emissivity is the ratio of a surface’s ability to emit radiation compared to that


blackbody surface under the same conditions. It is the single materials parameter
that governs radiative heat transfer. It is calculated based on Equation 1.5,

, , , ,
  (1.5)

Where

λ = wavelength [m-1]
ε(λ,ϕ,θ,T) = spectral directional emissivity [-], or the ratio of the object
emission to that of a blackbody at a given wavelength, temperature and angle
E(λ,T) = Planck Spectral Distribution of Emissive Power [W · m-3]

The Planck Spectral Distribution of Emissive Power is represented as

, (1.6)

Where,
3

h = Planck’s Constant = 6.63E-34 [J · sec]


c = Speed of Light = 3E8 [m · s-1]
k = Boltzmann’s Constant = 1.38E-23 [J · K-1]

Two or more mechanisms of heat transfer can occur simultaneously, of


course. For an example, we give two concentric cylinders having immobile, stagnant
and transparent gas between the two surfaces, as illustrated in Figure 1.

ε2 ,T2

r2
ε1 ,T1
r1

Figure 1.1. Two Concentric cylinders. The inner cylinder has temperature T1, radius
r1, and emissivity ε1. The outer cylinder has temperature T2, radius r2, and emissivity
ε2. The gas (tan color) between the two surfaces exhibits a thermal conductivity of k.

Here, the total heat transfer, Q, between the first cylinder surface onto the second
cylinder surface may be represented via Equation 1.7.

Radiation Conduction

  (1.7)
 
4

Where,
Q = heat transfer [W] from surface 1 to surface 2
L = Length of the cylinders [m]
k = conductivity of gas within gap

Conventionally, thermal radiation has often been neglected in engineering


heat transfer analyses since the Stefan-Boltzmann Constant is so small that such a
simplification is entirely reasonable under most practical circumstances. However,
this is no longer necessarily the case in the nuclear reactor environment, especially
given the current trajectory of modern nuclear reactor designs.

To achieve higher efficiencies, future nuclear reactor concepts are aimed at


operating temperatures significantly higher than current light water reactors (LWR).
At these high operating temperatures or during accidental high temperature
excursions in LWR, the dependence on thermal radiation for the expulsion of heat
from the system becomes critically important because of the fourth power
temperature dependence of radiated heat according to the Stefan-Boltzmann
relationship. The material property that dictates the extent of radiant heat is
emissivity, which is defined as the ratio of radiant heat from a material’s surface to
that of a black body. While rough approximations of emissivity have sufficed for light
water reactor heat transfer analysis for normal operating conditions, a stronger
understanding is necessary for heat transfer predictions of Gen IV high temperature
reactors or in the case of high temperature excursions in LWR. As a result,
experimental studies of emissivity coefficients for various relevant nuclear alloys
have been performed in recent years.[1][2][3][4][5][6]

The necessary foundation of experimental reference of reactor emissivity is


still under development. The emissivity evolutions of several key reactor alloys
through exposure conditions have not been investigated in great detail. For example,
we found no previous study on the emissivity of MSR alloys through corrosion from
molten FLiBe exposure. Granted, a single experimental study can only focus on so
5

many conditional variables at once. For nuclear reactor application, there exists a
variety of different alloys proposed to be used in a variety of different environments
and temperatures, varying between different reactor designs. A great many of the
circumstances have yet to be investigated.

Additional experimental investigation is a necessary first step. However, a


thorough and flexible mastery in any physical problem is achieved only when a
strong predictive capability grounded in theory is established. To this end, it is
desirable to develop a first-principles-based modeling for the case of reactor alloy
emissivity. However, emissivity modeling itself is also still a developing field. In order
to develop emissivity modeling of reactor alloys, it is necessary to develop the field
in general.

In short, this project focuses on the following goals:


1) To achieve a more detailed understanding of the evolution of the emissivity of
low-alloy steels under RPV conditions

2) To achieve a preliminary view into the emissivity evolution of several other


structural alloys under exposure conditions of Generation IV systems.

3) To advance current capabilities closer to a first-principles capability of


predicting spectral directional emissivity as well as total hemispherical
emissivity in nuclear reactor applications.

1.1 Reactor Pressure Vessel Alloys

Thermal radiation plays an especially pronounced role in expelling heat from


the reactor pressure vessel. For Pebble Bed Modular Reactors and Modular High
Temperature Gas Reactors during an accident scenario where the RPV reaches
300°C, it is expected that 80% of heat transfer in proposed RCCS designs will be
radiative[7]. In 2005, Ball et al. [8] studied the sensitivity in heat transfer simulations
of several transient scenarios for modular gas-cooled reactors. Several parameters
were adjusted during separate runs in this study including the parameter of RPV and
RCCS surface emissivity. The results of a Depressurized Loss of Forced Circulation
6

Accident for 600 MW(t) Prismatic Gas-turbine Modular Helium Reactor under slightly
different emissivity coefficients is given in Table 1.1.

Table 1.1. Scenario Evaluation for 600 MW(t) Prismatic Gas-Turbine Modular
Helium Reactor

Scenario Assumptions Max Fuel Temp [°C ] Max RPV Temp [°C ]
ε = .8 for RCCS
Depressurized surface and RPV 1494 555
Loss of surface
Forced
Circulation ε = .6 for RCCS
Accident surface and RPV 1508 609
surface

Low alloy steel SA508 and other similar steels such as SA-533 are currently
the industry standard for reactor pressure vessels and they remain as valid material
candidates for the RPV of many of the Gen IV designs. With that being said, there is
a potential ceiling to their application in future designs such as the non-modular
prismatic VHTR[9]. In this case, alloys with increased Chromium content such as
T91 or SA-336 are desirable candidates. It is important to understand the evolution
of the thermal emissivity coefficient of these materials under conditions
representative to their intended environment.

For reactor application, effects of roughness and oxide thin film interference
are of particular interest. Small scale modeling of emissivity coefficients in other
fields has focused on isolating these individual contributing effects[10][11]. In high
temperature reactor conditions, oxidation or some form of surface corrosion will
inevitably occur. An oxide layer on a metal substrate has different optical constants
compared to the metal substrate due to the dielectric nature of the oxide. An
oxidation layer can act as a thin film where diffraction occurs, thereby affecting
emissivity of the surface by changing the overall emission of the surface while also
inducing significant oscillations of spectral emissivity from interference
7

effects[12][13]. Such oscillations are visible in the work of Campo[14] in Figure 1.2
for bare iron and in the work of Cao[15] in Figure 1.3 for heavily oxidizing SA508.

Figure 1.2. Spectral emissivity progression across time for an iron sample oxidized
in air at 480oC. Constructive and destructive interference by oxide is visible.

Figure 1.3. Spectral emissivity progression of oxidizing SA508 at 500 °C (left) and
700 °C (right)
8

Surface roughness effects must also be considered since actual commercial


metal alloy surfaces are real surfaces with small cavities and surface features.
Surface features induce more surface contact interaction with propagated light,
which generally increases emissivity. It is well established that roughness has a
distinct impact on the emissivity coefficients of metals. However, the exact and
convenient description of this interaction across different materials and conditions is
still an open problem. In this study, systematically designed emissivity experiments
were conducted with actual reactor pressure vessel candidate materials to gain
insight into the effects of oxidation and surface roughness. The effect of temperature
on emissivity was also evaluated since the change of temperature since the change
of temperature can influence optical constants and different wavelength
contributions.

The influence of temperature is two-fold. Firstly, and most significantly,


temperature influences the Planck Distribution (Equation 1.6) which determines the
relative contribution of any given spectral range to total emission. Figure 1.4
presents the Planck Distribution for several temperatures as well as the Cumulative
Distribution Function for Planck. As temperatures rise, a greater preference towards
the shorter wavelengths occurs.
9

Planck Distribution [W·m-2·μm-1]


4000
3500
3000
T = 500°C
2500
E(λ,T)

T = 300°C
2000
1500 T = 100°C

1000
500
0
0 5 10 15 20 25
Wavelength [μm]

Figure 1.4. Planck Distribution of Emissive Power. As temperatures increase,


Emissive power increases across all wavelengths, with greater relative increases
occurring at lower wavelengths.

0.8
Cumulative Density of E

0.6 T = 500°C
T = 300°C

0.4 T = 100°C

0.2

0
0 5 10 15 20 25
Wavelength [μm]

Figure 1.5. Cumulative Distribution Function of E(λ,T). At greater temperatures, the


proportion of energy emitted at shorter wavelengths increases.

Temperature-induced shifts of the Planck Distribution are referred to as


Planck shifts in this study (for convenient shorthand) and while they do not effect
10

spectral emissivity, they do influence total emissivity as long as spectral emissivity is


non-uniform across wavelength. Secondly, Spectral emissivity can experience minor
changes associated with the temperature dependency in the optical constants. As
shorthand, temperature-induced changes in the optical constants of a material are
referred to optical shifts.

1.2 Structural Materials

While emissivity of the RPV surface has received some attention, it has also
been shown that radiative heat transfer can be very significant in other spaces. For
instance, Abrosek et al. found that radiation has a significant effect in molten salt
applications having high temperatures at or above T =1123 °K or under laminar flow
conditions (Re<500) for pipes having a diameter of 1 cm or greater.[16] It is
expected that thermal radiation will be very significant within a VHTR core. Core-to-
Pressure Vessel heat transfer of a scaled loop experiment was simulated by Hong et
al. for a 270 kW helium heater element where transfer was simulated with radiation
neglected and with radiation included.[17] The results are provided in Figure 1.6.

Figure 1.6. Heat transfer simulations performed by Hong et al. for a 270kW Helium
heater scaled experiment. The core in the radiation-neglected case reaches a
temperature 500 °C greater than that of the radiation-included simulation.
11

The MSBR project performed initial calculations to evaluate heat transfer of


heat exchangers of the MSBR during the afterheat from a reactor shutdown.[18]
The emissivity coefficients of the shell surfaces within the heat exchangers were
varied in these calculations. For the 281-MW heat exchanger, a coefficient of 0.1
would yield an inner shell temperature of 1319 °C while a coefficient of 0.2 would
yield an inner shell temperature of 1135 °C, a difference of 184 °C. Similar
differences were found for higher power exchangers during steady state operation.

Despite the significance of emissivity within the Gen IV reactor environment,


there exists little to no prior work on the evolution of surface emissivity as these
alloys are exposed to Gen IV environments of liquid sodium, molten salt, helium gas,
and supercritical CO2. Air oxidation studies have been performed[5][19], but not
exposure studies within the actual intended media - most likely because of the
infrastructure necessary to perform such exposures.

Much of the same effects expected to occur with the low alloy steel RPV
surfaces are expected to happen with other structural alloys in other applications
spaces for various other reactor designs. For example sCO2 often gives a harsh,
oxidizing environment and so if stainless steel is used in a sCO2 heat exchanger,
there may be oxide growth similar to the RPV low alloy steel. For stainless steel in
an environment such as liquid sodium, it is not expected that much oxidation will
occur. However, the surface may still roughen a bit. Ultimately, it is necessary to
collect initial data on the emission of alloys exposed to conditions closer to their
intended application space. Although a detailed parametric study for these alloys in
these environments was not possible due to infrastructure constraints (one or two
samples per environment), valuable initial data was still acquired.

1.3 Experimental Techniques

There exists two practical general methods of experimentally determining


thermal emissivity. The first approach, also known as the calorimetric
12

method[20][21][22], involves a delicate accounting of heat transfer. Usually this


approach requires isolating a sample such that the mechanisms of convection and
conduction are minimized and then evaluating the overall heat transfer out of the
sample and through the experimental system. Through energy balance, the
emissivity coefficient of the sample can be obtained based on the amount of radiated
energy. The second approach, the radiometric method, involves optical
measurement of the object. This approach can be broken up into two subsets of
measurement, one involving the radiated emission of the object and the other
involving the reflectance of the object combined with the application of Kirchhoff’s
Law. Measurements may be taken using a pyrometer[23][24][25], spectroscope[26]
or a Fourier Transform Infrared Spectrometer.[27][28]

Unless laser heating is used[29], the calorimetric method is generally limited


to the total hemispherical emissivity coefficient: it cannot readily be used to
determine spectral or directional values. Furthermore, there exists several
complications in this approach. To minimize convection, the measurement system
should maintain a good vacuum. Unless many resources are devoted to suspending
the sample in the vacuum, contact of the sample with some surface and some
resultant conduction is unavoidable. Thermocouples used to monitor temperatures in
the system may also complicate the balance. To maximize the contribution of
radiation to overall heat transfer, the sample should be maintained at high
temperatures which may make maintaining the vacuum more difficult. Lastly, the
radiative properties of the material enclosing the sample should ideally be known
ahead of time.

The method of directly measuring emitted light from an object through


radiometry is more flexible than the calorimetric method. An FTIR is often used for
this purpose. Under ideal circumstances, direct emission measurement
methodologies generally follow the relation,

,
,   (1.8)
,
13

Where P is spectral emissive power of the measured sample per unit area and
under ideal circumstances,

,   , , (1.9)
Perfect blackbodies do not exist in nature and a reference having a known
emittance, ε(λ,T) is used. Actual experimental methods typically follow

,
,    , (1.10)
,

Where the P’s are the measured parameter and εR is known ahead of time, usually a
relatively high value to limit uncertainty.

There exist several complications for this measurement approach. Firstly,


oxidation may occur during measurement at elevated temperatures thus
contaminating the data. Secondly, absorption from ambient oxygen and water
molecules may occur at points along the measured wavelengths. Thirdly in our
particular case, temperature variations between the samples may occur based on
the material’s thermal resistance and affixation to the vertically-inclined stage. Of the
four experimental methodologies utilized in this piece, one methodology follows this
approach.

Radiometric reflectance measurement techniques that utilize Kirchhoff’s Law


(Equation 1.11) avoid some of the complications found with the direct emission
measurement path.

, ,   , , (1.11)
Unlike emission measurement, reflectance measurements do not require
elevated temperature to generate adequate signal from thermal radiation and so the
spectral range of measurement isn’t bounded by the Planck Distribution. After
calibration, only measurement of one object (the sample) is required to generate
data – although usually a reflective sample (often a gold surface) is also used for
calibration. Issues of in-situ oxidation and surface area variation are no longer
concerns, either. However, additional lengths must be taken to ensure proper
14

measurement if the sample surface is rough. If light scatters diffusely rather than
specularly, all light must still be collected and accounted for. To ensure the collection
of all light, an integrating sphere or other techniques may be used. Of the four
experimental methodologies utilized in this piece, three follow this approach.

1.4 Modeling Efforts

Emissivity in the nuclear environment presents a complicated modeling


problem. The brunt of emissivity modeling capability lies on the modeling of ideal,
smooth, unoxidized surfaces at a small range of temperature. Every deviation from
the ideal complicates the prompt. Of course, the nuclear environment features
materials far from the ideal. Alloys may oxidize, change composition, roughen, and
experience a wide range of temperatures through operation. Furthermore, similar to
Equation 1.7 and Figure 1.1, the total hemispherical emissivity coefficient is what is
most commonly used as an input in nuclear heat transport models while most
emissivity modeling work has focused on spectral directional models. Current
emissivity modeling capabilities fall short of addressing nuclear material emissivity
and so it’s necessary to advance modeling as a whole accordingly.

Spectral directional emissivity, which describes emissivity at particular


wavelengths and directions, depends on a variety of factors in addition to
wavelength and direction. We mention material composition, surface oxidation,
surface roughness, and surface temperature, among others. Total hemispherical
emissivity is an integral of the spectral emissivity over all directions and all
wavelengths. For accurate evaluation of heat transfer at very high temperatures, it
would be ideal if total hemispherical emittance could be rapidly obtained from a first-
principles based approach (most likely grounded in Electromagnetic Theory and
Maxwell’s Equations) where all of the dependencies mentioned above are
accounted for. This would extend predictive capability far beyond that of the limited
sources of existing experimental literature.
15

Total hemispherical emittance can be obtained from experimental data or


from experimentally observed trends and relations [30][31]. In some cases, theory is
indeed applied to culminate into a total hemispherical emittance such as has been
done by Rosenberg et. al[32], however, this is limited to the case of idealized, flat,
unoxidized surfaces. Despite the importance of total hemispherical emissivity in high
temperature systems, theoretical models describing the effects of roughness,
oxidation, etc., have yet to be developed.

Greater progress in the application of theory, however, has been made


regarding spectral and spectral directional emissivity. First, the acquisition of
experimental data is relatively easier than in the case of total hemispherical
emissivity since the total values require the calorimetric method or many radiometric
measurements whereas a spectral directional value requires a single radiometric
measurement. Not only does Kirchhoff’s Law enable the radiometric reflectance
measurement technique, it also enables modeling - a large body of theoretical work
dedicated to spectral directional reflectance is immediately transferrable to spectral
directional emissivity. While challenges remain, many techniques and methods have
been developed and are still being refined in modeling spectral directional emissivity.
Various approaches are being taken to resolve effects such as temperature
dependence, roughness and surface oxidation.

As mentioned previously, the effect of temperature on total emittance is two-


fold through Planck Shifts and Optical Shifts. While Planck Shifts are the most
immediate influence that temperature has on total emittance, optical shifts may also
influence the spectral directional emittance and thereby the total emittance. Several
theoretical models have been developed to describe the optical constants within
certain ranges of wavelength, material, and temperature. A few being the Hagen-
Rubens Relation[33], Drude Model[34], Lorentz Model, Lorentz-Drude Model[35],
Brendel-Bormann Model[36], Drude-Roberts[37] and the dampened Drude model
developed by Chen and Ge[38]. Models such as Lorentz and Brendel-Bormann are
generally better suited to dielectric materials whereas the Drude Models and the
16

Hagen-Rubens Relationship (which follows the Drude Model) are more suited to
metals. Additional measures beyond the Drude model must be taken in order to
describe the anomalous skin region[39] or the interband absorption in noble
metals[40]. Critical reviews of different models of optical constant for different
application spaces and ranges of validity although the most general of expression of
optical constants are available in open literatures [41].

For a roughened surface, Wen and Mudawar [10] demonstrated how simple
roughness factors developed by Agababov[42][43][44] could be implemented to
achieve good experimental agreement for spectral emissivity of aluminum surfaces.
Theoretically in-depth methods have also been developed focusing on the
interaction of light with the surface. These methods would include using
electromagnetic theory and Maxwell’s Equations or ray-tracing methods to evaluate
some input surface profile. Unfortunately, the field of surface metrology is still open
and there is not yet a generally-agreed upon method of characterizing and
mathematically representing a real surface. However, modeling efforts have found
decent success by utilizing pseudorandomly-generated Gaussian-distributed surface
profiles [45][46]. Generally, a rigorous method of generating spectral directional
emissivity for a non-ideal (roughened, patterned, etc.) input surface would be the
direct solving of Maxwell’s Equations through means such as the Finite-Difference
Time-Domain method, Finite Elements, or Rigorous Coupled Wave Analysis. There
exist several prewritten scripts and software programs that implement one of these
techniques to solving Maxwell’s Equations, for example, MEEP[47] and Georgia
Tech’s RCWA[48], have been used for this very purpose. However, direct solving
methods require significant computational resources compared to simpler methods
such as ray-tracing. This is especially true for pseudorandom surface profiles as
opposed to repeating structures (structures such as grating geometries or smooth
cavities) if meshing is to be considered for a method such as Finite Elements.

For a rough surface, the region of physical validity of the Geometric Optics
Approximation (GOA, also referred to as Geometric Optics, Geometric Optics Ray-
17

tracing, Ray-tracing, and GO) has been studied by various authors. The GOA
approach saves computational resources by tracing the interaction of ray of light with
surface instead of solving Electromagnetic and Maxwell’s equations. Tang and
Dimenna identified several parameters having strong influence on the accuracy of
the GOA method, namely incident angle, ratio of rms deviation to correlation length
and the ratio of the rms deviation to wavelength[49] and produced a regime map
based on the latter two parameters. Fu and Hsu identified additional dependency of
incident angle and produced an updated regime of validity including all three
parameters[50]. Both of these studies focused on one-dimensional roughness. As
computational capabilities have increased, so has the ability to model more complex
geometries, such as two-dimensional surfaces. Both Tang[51] and Bergstrӧm et al
[52] investigated the ray tracing of 2D Gaussian random rough surfaces and found
minor differences compared to a similar models of 1D geometry. While the GOA
method is limited at high incident angles, recent work has shown the capability of the
RCWA method to match experimental data in such conditions.[53]

The addition of a thin dielectric layer (an oxide in the nuclear case) can add
further complication by introducing interference effects. Several authors have found
reasonably effective ways of dealing with this as well. Iuchi modeled emissivity of
oxidized metal by treating the oxide layer and the metal substrate as parallel, ideal,
layers[11]. This method was capable of including interference effects which can be
particularly distinctive in the example of iron oxidation[14]. The method of Phase ray-
tracing attempts to include interference effects within the framework of the GOA
method. This method was used by Lee et al. [54] to model gold and silicon oxide
coatings on a silicon substrate. Further evaluation of the method identified limitations
and the necessity to normalize associated calculations due to the violation of energy
conservation[55]. More recently, a similar method, dubbed Hybrid Partial Coherence
and Geometric Optics (or HPCGO), was developed by Qiu and Wu and compared to
Maxwell equation solving methods[56]. Zhang and Zhao[57] included interference
effects in their GOA adaptation, although the interference effects included in their
18

model address multiple reflections on the surface of a single layer rather than the
interference effects associated with layers.

As discussed, there are a variety of physics-based techniques that can be


used in the attempt to resolve several dependencies in the formulation of a spectral
directional emissivity. In general, the models previously discussed produce discrete
spectral directional emittance (or reflectance) values or they produce a bi-directional
reflectance distribution function (BRDF) which can then in turn be integrated to give
discrete values of spectral directional emittance.

This study presents the simple concept of generating a total hemispherical


emittance coefficient by iteratively running a spectral directional emittance model for
directions and wavelengths representative of what would be expected in nature. The
spectral directional emittance model accounts for dependencies of wavelength,
direction, composition, temperature and roughness and so the total hemispherical
emissivity captures the desired dependencies of composition, temperature and
roughness as well. While the spectral directional model presented in this paper is by
no means the most-advanced or theoretically fundamental, it still demonstrates that
the accounting of dependencies at the spectral directional level can propagate
through to a total hemispherical emissivity. If first-prinicples are to be applied for the
emissivity modeling of nuclear alloys, this is a gap that will most likely need to be
filled.

Of the many modeling techniques reviewed, the GOA was chosen for several
reasons. First, it is a flexible, straightforward technique that can be internally
authored and altered. This way, it could easily be written to be looped for Monte
Carlo sampling purposes. Secondly, it is not particularly computationally intensive
and so the sampling method could be explored with quicker turnaround. Thirdly,
Monte Carlo ray-tracing methods are already commonly used to model radiative
heat transfer in complex systems that may not be solvable deterministically. It was
expected that the established method of sampling energy bundles[58] could be
19

applied analogously to spectral directional models, such as the GOA, to derive a


total hemispherical emissivity.

2 Experimental Methods 

2.1 Test Materials

2.1.1 Low Alloy Steels

Several T91 and SA508 steel (alloys for RPV) samples were EDM-cut and
mirror-polished before being subjected to either air oxidation or roughening. The
compositions of the materials used are given in table 1. The oxidation temperatures
were selected based on code standards, postulated transient scenarios, and on
expected reaction thresholds.

Table 2.1. Nominal Compositions of the Low Alloy Steels Tested

Alloy Fe Cr Ni Co Mo Al Mn Si C

SA508 Balance .17 .17 - .48 - 1.28 - .27 max


T91 Balance 8.37 .21 - .9 .01 .45 .28 .14 max

SA508 steel is allowed in the ASME Boiler and Pressure Vessel Code for
indefinite nuclear service up to 371 °C.[59] However, most operating temperatures of
PWR and LWR reactor pressure vessels are roughly 290 +/- 30 °C. [60] It is
unrealistic to try and expose samples to operating temperatures at durations typical
of pressure vessel lifetime. For this reason, an elevated temperature of 350 °C was
chosen to accelerate oxide growth at durations up to 200 hours. To generate a
upper bound on oxidation, another sample was exposed to a much-higher
temperature of 600 °C for 5 hr. Should the allowable temperature ranges of SA508
be extended in future code standards, this environment still easily surpasses the
upper bound of relevant temperatures – this well exceeds the simulations performed
20

by Ball[8] presented previously in Table 1.1. The 600 °C sample represents an


excessively oxidized RPV.

Due to the increased chromium content in T91, it was expected that only very
thin oxide layers would form during high temperature air exposure and that the
change in emissivity coefficients would be relatively limited. Oxidation experiments
started at 600°C for 3 hours but initial SEM cross section analysis showed negligible
change in surface condition after this exposure. An extreme temperature of 750°C
was chosen to try and determine measurable emissivity changes from surface
oxidation. A General Signal Lindberg Box Furnace was used for the air oxidation
experiments (Figure 2.1).

Figure 2.1. Lindberg furnace used to oxidize samples in high temperature ambient
air environments.

Some samples were also roughened to investigate potential effects of surface


roughness (Table 2.2). Standard SiC grinding disks were used for this purpose.
21

Shot-peened samples were subjected to stainless steel particles at approximately 40


psi until the sample surface appeared uniformly roughened upon visual inspection.

Table 2.2. Applications, Test Materials and Conditions


Test Material Intended Effects Test Conditions
300C, 100 hr
350°C, 10 hr
Oxidation during Steady 350°C, 50 hr
State Operation 350°C, 100 hr
350°C, 150 hr
SA508 350°C, 200 hr
Oxidation Limit 600°C, 5 hr
800-grit finish
320-grit finish
Roughness Effects
120-grit finish
Shot-peening
Oxidation Limit 750°C, 5 hr
T91 320-grit finish
Roughness Effects
Shot-peening

A ZYGO white light interferometer was used to measure surface profiles for
each sample used in the roughness sections of the study (Table 2.3). For each
sample, three profiles were measured and each value was averaged between the
three profile quantities (max peak-to-valley value, root-mean-squared roughness
value, arithmetic average). Values given in parentheses are the standard deviation
within the averaged set.

Table 2.3. Profilometry measurements of unoxidized samples


Test Material Test Conditions PV [μm] RMS [μm] Ra [μm]
Mirror-finish 1.932 .0313 .016
reference (.3464) (.0123) (.0017)
SA508 .936 .017 .0117
800-grit finish
(.5167) (.013) (.0072)
320-grit finish 2.826 .134 .0963
22

(.4392) (.016) (.0105)


3.403 .2933 .232
120-grit finish
(.6239) (.0929) (.072)
20.254 2.544 2.013
Shot-peen
(1.801) (.2682) (.2428)
Mirror-finish 1.347 .0147 .0077
reference (.7273) (.0055) (.0031)
2.662 .0867 .060
T91 320-grit finish
(.5527) (.0235) (.01637)
16.551 2.192 1.745
Shot-peen
(.8294) (.2675) (.2320)

A few cross-section SEM images were taken on the air-oxidized samples


using a Jeol 6610 SEM. The oxide layer that formed on the SA508 was very fragile
and appeared to partially detach from the surface during cross section polishing. The
cross section of the surface of the sample oxidized at 350 °C over 10 hours is
presented in Figure 2.2 while the heavily oxidized 600 °C, 5 hour SA508 sample is
provided in Figure 2.3. T91 formed a decently thick oxide, more secure that the
SA508, shown in Figure 2.4.
23

Figure 2.2. SEM cross section image of SA508 oxidized at 350 °C over 10 hours.
Red arrows indicate the oxide layer. Detachment appears to have occurred along
the right side of the surface.

Figure 2.3. SEM cross section image of SA508 oxidized at 600 °C over 5 hours. Red
arrow indicates the oxide layer. Detachment appears to have occurred through most
of the oxide layer.
24

Figure 2.4. SEM cross section image of T91 oxidized at 750 °C over 5 hours. Red
arrow indicates the oxide layer.

This study focused on four environments of interest and one or two candidate
alloys that are being considered for each environment. The four different exposure
environments were helium gas [61] (emulating a HTGR core), molten salt [62]
(molten salt reactor (MSR) components), liquid sodium [63] (sodium fast reactors
(SFR) cladding and structure), and supercritical CO2 [64] (supercritical CO2 heat
exchangers that are being considered for high temperature advanced reactor
concepts). The temperatures of these environments were selected based on
proposed steady state operating conditions for the respective reactor types. All
samples were exposed in a flowing medium except for the molten salt experiments.
The sample information and exposure parameters are provided in Table 1. Impurity
measurement is more difficult for molten salt than the other environments studied
and molten salt impurity content could not be attained. For the molten salt
environments, hydrofluorination was used to purify the salt.

Table 2.4 Summary of test materials and exposure environments and conditions
used to create samples for emissivity testing.
Reactor Temperature,
Test material Environment
Application duration
Haynes 230 - -
Reference Inconel 617 - -
samples Hastelloy N - -
SS-316 - -
He gas, carburizing
HTGR core <0.5 ppm H2O, 10 ppm CO, 100 1000°C, 500 hr
Haynes 230
structure ppm CH4
He gas, oxidizing 1000°C, 500 hr
25

4 ppm H2O, 40 ppm CO, 20 ppm


CH4
He gas, (carburizing)
<0.5 ppm H2O, 10 ppm CO, 100 1000°C, 500 hr
ppm CH4
Inconel 617
He gas, oxidizing
4 ppm H2O, 40 ppm CO, 20 ppm 1000°C, 500 hr
CH4
Molten LiF-KF-NaF
Hastelloy N 850°C, 1000 hr
MSR core No flow
structure Molten LiF-BeF2
SS-316 750°C, 1000 hr
No flow
SFR Liquid sodium
cladding SS-316 10 ppm O2 650°C, 1000 hr
& structure Flow rate ~ 6 m/s
Research grade CO2
H2O < 3ppm, N2 < 5ppm, THC
Haynes 230 < 1ppm, Ar+O2+CO < 1ppm 650°C, 400 hr
Advanced
Flow rate ~ 0.11 kg/hr
reactor
20 MPa
sCO2
Research grade CO2
heat
H2O < 3ppm, N2 < 5ppm, THC
exchanger
SS-316 < 1ppm, Ar+O2+CO < 1ppm 650°C, 400 hr
Flow rate ~ 0.11 kg/hr
20 MPa

Four alloys were tested in total, with their nominal compositions presented in
Tables 1.5 and 1.6. Profilometry was performed on all samples and measurements
are provided in Table 1.7 RMS gives the root mean squared of profile heights, Ra
26

gives the arithmetic average of absolute values of the surface profile. PV, or Peak-
to-Valley, gives the maximum peak-to-valley difference in the measured profile.

Table 2.4. Nominal Compositions of the Ni-based Alloys Tested

Alloy Fe Cr Ni Co Mo Al Mn Si C W Ti
Hastelloy-N <5 7 Bal. <0.2 16 <0.35 <0.8 <1 <.08 0.5 <0.35
Haynes 230 <3 22 Bal. <5 2 0.3 0.5 0.4 0.1 14 <0.1
Inconel 617 1.6 22.2 Bal. 11.6 8.6 1.1 0.1 0.1 0.05 - 0.4

Table 2.5. Nominal Compositions of the Fe-based Alloy Tested

Alloy Fe Cr Ni Cu N P S Mo Mn Si C
316
Stainless Bal. 16.83 10.03 0.38 0.05 0.03 <0.01 2.01 1.53 0.31 .0225
Steel

Table 2.6 Profilometry measurements of exposed samples

Sample PV [μm] RMS [μm] Ra [μm]


Haynes 230,
14.015 1.856 1.515
Carburizing
(0.8387) (0.016) (0.009)
Helium
Haynes 230,
12.324 0.941 0.730
Oxidizing
(1.147) (0.112) (0.086)
Helium
Inconel 617,
14.020 1.705 1.424
Carburizing
(0.059) (0.095) (0.097)
Helium
Inconel 617,
16.238 1.632 1.24
Oxidizing
(2.352) (0.139) (0.068)
Helium
Hastelloy-N, 12.281 1.499 1.273
FLiNaK (1.086) (0.162) (0.160)
316
3.352 0.198 0.145
Stainless
(0.279) (0.002) (0.003)
Steel, FLiBe
27

316
6.894 0.323 0.241
Stainless
(1.293) (0.010) (0.003)
Steel, Na
Haynes 230, 4.806 0.402 0.260
sCO2 (0.534) (0.081) (0.074)
316
10.257 1.624 1.234
Stainless
(0.196) (0.535) (0.470)
Steel, sCO2

2.1.1.1 High Temperature He Gas


Nickel-based alloys Inconel 617 and Haynes 230 are the leading candidate
metallic alloys for the construction of high temperature gas-cooled reactor (HTGR)
[65], where it is likely that helium will be used in the primary loop of the reactor, and
outlet temperatures may reach as high as 950 °C during steady state operation. In
this work, samples were exposed in Idaho National Laboratory’s high-temperature
helium gas flow loop at 1000 °C for 500 hours, [66] with exposure conditions
designed to result in either oxidation or carburization of the alloys. The oxidative
environment contained helium with impurity levels of 4 ppm H2O, 40 ppm CO and
20 ppm CH4, while the carburizing environment contained helium with impurities
<0.5 ppm H2O, 10 ppm CO and 100 ppm CH4.

2.1.1.2 Molten Salt


Molten salt reactors are expected to operate at temperatures of 700 °C or
higher.[67] We studied Hastelloy N exposed to high-purity molten LiF-NaF-KF
(FLiNaK) at 850 °C for 1000 hours in a 316 stainless steel crucible, which was
procured from an earlier study at the University of Wisconsin by Sellers [68].
Exposure apparatus is presented in Figure 2.5.
28

Figure 2.5. Inconel box furnace used for FLiNaK molten salt corrosion tests.

Additionally, a 316 stainless steel sample exposed to molten LiF-BeF2 (FLiBe)


for 1000 hours at 700 °C in a 316 stainless steel crucible was also investigated. This
sample was also procured from a recently concluded project at the University of
Wisconsin [69]. 316 stainless steel has emerged as a lead candidate for the
construction of the vessel in fluoride salt-cooled high temperature reactor (FHR).
FLiBe is being considered as the primary coolant for the FHR for non-fueled
coolants, and it also forms the basis for fuel-dissolved salts [62].

Originally, two FLiBe-exposed stainless steel 316 samples were procured for
this study from the experimental work of Zheng et al.[69]. Both samples had been
exposed for 1000 hours at 750°C but one had been exposed in a 316 stainless steel
crucible while the other had been exposed in a graphite crucible (setup in Figure
2.6). Each sample was too small to be measured through the direct emittance
system and so measurement via reflectance was pursued. The graphite-enclosed
sample had a rougher surface profile, likely due to accelerated corrosion, which
29

bordered on the valid measurement range of the VERTEX reflectance measurement


system and is not presented in this piece.

Figure 2.6. Atmosphere-controlled glove box housing FLiBe corrosion test.

2.1.1.3 Liquid Sodium


Stainless steel 316 is being considered for use in sodium-cooled fast reactors
which are expected to operate at temperatures as high as 700 °C [70]. Samples for
this study were exposed to liquid sodium at 650 °C in a sodium loop facility (Figure
2.7) for 1000 hours at the University of Wisconsin [71]. The oxygen concentration in
the liquid sodium was 10 ppm, and the sodium flow velocity was 6 m/s. A small
decrease in mass (<0.03% of total mass) was observed after exposure.
30

Figure 2.7. Liquid sodium loop testing facility for corrosion testing of samples used in
this study.

2.1.1.4 Supercritical CO2


Samples of 316 stainless steel and Haynes 230 were laser-sectioned from
virgin sheets, cut via electrical discharge milling, ground with 800 grit silicon carbide
(SiC) paper, and exposed to supercritical sCO2 in a specially designed autoclave
system [72].

This environment was selected because of the increasing interest in the sCO2
Brayton cycle, which provides higher efficiencies than the steam Rankine cycle at
650 °C and above [73] and is therefore deemed to be attractive for advanced reactor
concepts. For this study, the samples were exposed to 650 °C sCO2 at a pressure
of 20 MPa. A system-average flow rate of 0.11 kg/hr was maintained, resulting in a
CO2 refresh rate every two hours. H2O, N2, and O2 impurities were monitored and
regulated below a few ppm. A photograph of the SC-CO2 autoclave test facility used
in this study is shown in Figure 2.8.
31

Figure 2.8. Photograph of the SC-CO2 autoclave system used in this study.

2.2 Measurement Apparatus

Four different measurement methodologies were explored in this project, with


each having individual capabilities and limitations.

2.2.1 ET100 Emissometer (Reflective Radiometry)

A Surface Optics ET-100 Emissometer (Figures 2.9 and 2.10) was used to
conduct spectral directional reflectivity measurements at incident angles of 20 and
60 degrees. Through the use of an integrating sphere, both specular and diffuse
reflecting light were measured after initial incidence. Spectral directional emissivity
having an emission angle equal to the measurement incidence angle could then
readily be determined using Kirchoff’s Law. Measurements were taken at room
temperature. The ET-100 features six spectral ranges from 1.5 to 21 μm. Their
respective ranges are provided in Table 1.8.
32

Figure 2.9. Surface Optics ET-100 Emissometer in carrying case.

Figure 2.10. Surface Optics ET-100 Emissometer connected to desktop


measurement software. Samples are placed on top of the uncovered golden
integrating sphere (currently covered with cap in the image) and then measured.
33

Table 2.7. ET-100 Incident Angles and Spectral Ranges


Incident Angle Spectral Band Range
λ = 1.5 to 2.0 μm
λ = 2.0 to 3.5 μm
λ = 3.0 to 4.0 μm

θ = 20° λ = 4.0 to 5.0 μm


λ = 5.0 to 10.5 μm

λ = 10.5 to 21.0 μm

λ = 1.5 to 2.0 μm
λ = 2.0 to 3.5 μm

λ = 3.0 to 4.0 μm
θ = 60°
λ = 4.0 to 5.0 μm
λ = 5.0 to 10.5 μm

λ = 10.5 to 21.0 μm

As shown in Table 1.8, a total of twelve spectral directional values can be


measured. In this study, each data point was measured multiple times and the
average data point value is presented.

The spot size of the ET-100 is limited to a diameter of approximately 15 mm.


Unfortunately, due to sample size restrictions in the molten salt, liquid sodium, and
high temperature helium gas environments, the samples from these tests were not
large enough to occupy the entire spot size diameter. For convenience, this
methodology is simply referred to as ET-100 in this piece.

2.2.2 SOC-100 HDR w/ Nicolet FTIR (Reflective Radiometry)

The Surface Optics SOC-100 HDR used in this project (Figure 2.11) was
managed by the Concentrated Solar Power Group at the National Renewable
Energy Laboratory in Golden, CO. The device used a heated stage which was used
to take elevated temperature measurements on the RPV samples while under a
nitrogen purge. Measurements could be taken at several separate angles and, if
34

desired, at two light polarizations. The apparatus featured a dome-like measurement


space which enabled hemispherical-directional reflection measurement rather than
directional-directional reflection measurement (as was the case with the Bruker).
This ensured measurement of both diffusely and specularly reflecting light. A
reflective gold reference sample had to be used for calibration before each
measurement. The measurement spot size of this device was also about 15mm in
diameter which was unfortunately too large for several of the structural alloy
samples. The spectral range of the device ranged from 1.7 to 25.42 micrometers
which gave a very wide range considering the temperatures of reactor operation.
This measurement system was the go-to for the low alloy steel samples. For
convenience, this system is referred to as the NREL FTIR in the rest of this piece.

Figure 2.11. SOC-100 Hemispherical Directional Reflectometer Coupled to Nicolet


FTIR, managed by NREL Concentrating Solar Power Group
35

2.2.3 Bruker VERTEX 70 FTIR with Hyperion Microscope (Reflective


Radiometry)

The Bruker VERTEX 70 apparatus used in this project (Figure 2.12) was
managed by Kats Group from the Electrical and Computer Engineering Department
at UW-Madison. While a heated stage could be used in conjunction with the device,
all presented measurements produced from this methodology were taken at room
temperature. The device featured a wavelength range of 1.25 to 17 micrometers and
data was collected from specularly-reflected light from a cone of incidence ranging
from approximately 5-23 degrees. This methodology featured the smallest spot size
(~100µm) of all systems and no sample was too small to measure. This
methodology is referred to as Bruker reflectance.

Figure 2.12. Bruker VERTEX 70 FTIR coupled with Hyperion microscope.


36

2.2.4 Bruker VERTEX 70 FTIR (Emissive Radiometry)

The Bruker VERTEX 70 was used again, this time to measure emission
directly with a liquid-nitrogen-cooled mercury-cadmium-telluride (MCT) detector. A
vertical heated stage was used to heat samples up to for each measurement. Since
the quality of data is ultimately defined by radiant heat flux, it was necessary to heat
the sample up so that it emitted more energy than the surrounding background. A
schematic is provided in Figure 2.13.

Figure 2.13. Schematic of VERTEX 70 emission measurement methodology.

During emission measurement, samples were vertically affixed to a heated stage


using Kapton tape. Measurements were taken at normal incidence at temperatures
up to 300 °C. Here we assume the sample emission signal S measured by our FTIR
can be decomposed as:

,   , , (2.1)

= System response [arb. units]


37

Eqn. 2.1 is an approximation that ignores the contribution of thermal emission


from the cooled detector and the room-temperature components of the instrument.
We found that these contributions do not significantly affect the resulting
measurement for sample temperatures of 300 °C (see supplementary section [74]),
so this temperature was used for all measurements. By comparing the signal from
our samples to that of a blackbody reference at the same temperature, SR(λ,T),
which has a known emissivity εR(λ,T), we extracted the sample emissivity as:

,
,   , (2.2)
,

The blackbody reference was a silicon wafer with a coating of vertically


aligned carbon nanotubes, with an emissivity, εR(λ,T), of .90 +/- .02 that was
assumed to not change significantly with wavelength. All measurements were
performed at normal incidence and so angular dependence is not considered. A
detailed discussion on error and uncertainty in our methodology is provided in our
supplementary section [74].

During emission measurement, samples were vertically affixed to a heated


stage using Kapton tape, and a highly conductive thermal grease (Timtronics Red
Ice 611 HTC) was used to fill the contact gap. Measurements were taken at near-
normal incidence with an approximate numerical aperture of 0.003.

In this setup, the measurable wavelength range is not actually clearly defined.
Several calculations of the Planck Distribution were performed to estimate valid
wavelength ranges based on sample temperature and the temperature of the
surroundings. A quick comparison is given in Figure 2.14.
38

1
E(λ,T) / Emax(λ,T=300C)

0.1

T = 300 °C
T = 200 °C
0.01
T = 40 °C

0.001
0 5 10 15 20 25
Wavelength [μm]

Figure 2.14. Normalized Planck Distributions.

It is apparent that for any temperature, signal will eventually plummet as


wavelengths get much shorter, leading to large noise and uncertainty. At the much
longer wavelengths, in addition to the distribution decreasing, it is apparent that the
signal of the surroundings will start to compete with the signal from the sample
resulting in noise and uncertainty. As can be seen in Figure 2.14 above, while λ < 5
μm, E(λ,T=300°C) is roughly two or more magnitudes greater than E(λ,T=40°C) but
while λ > 15 μm, E(λ,T=300°C) is less than one order of magnitude greater than
E(λ,T=40°C). Initial measurement of a reference Haynes 230 sample yielded this
pattern (Figure 2.15.)
39

1 1
ε, T = 200 °C
0.9 ε, T = 300 °C
E/Emax,  T = 200 °C
0.8
E/Emax,  T = 300 °C
0.7

E(λ,T)/Emax(T=300C)
Emissivity Coefficient

0.6

0.5 0.1

0.4

0.3

0.2

0.1

0 0.01
0 5 10 15 20 25
Wavelength [μm]

Figure 2.15. Preliminary emissivity measurements of Haynes-230 reference sample


superimposed w/ normalized Planck distribution functions. Noise develops when
Planck grows weaker and also when background competes (longer λ).

After inspecting the preliminary results, it was found that decent data could be
obtained in a wavelength range between 3<λ<18 μm while the sample surface
maintained a temperature of roughly T = 300 °C. This is the range that is used
throughout the emission measurements. Measurements at lower temperatures
resulted in smaller measurement ranges.

The spot size of the direct emission methodology was the second-smallest of
the experimental methodologies. Every sample except for the FLiBe – exposed
samples could fill the spot. However, the SA 508 samples could not be reliably
measured using this methodology due to temperature constraints. Samples had to
maintain an elevated temperature of 300 °C during measurement and the SA508
40

readily oxidized from this requirement –even from brief durations at this temperature.
This methodology is referred to as Bruker emittance.

2.3 Sources of Error and Uncertainty

The ET-100 was delivered as an open-and-go system with straightforward


and quick procedures already developed by Surface Optics Corporation. Deviations
between consecutive measurements were often negligible. The NREL FTIR system
had a similar circumstance with the added benefit of experienced managing staff.
For both of these systems, error and uncertainty are assumed minimal and any
reported deviation is simply the standard deviation between consecutive
measurements for the same sample. However, due to spot size restrictions many of
the samples could not be adequately measured on the established devices. The
Bruker reflectance and emission methodologies had to be developed from the
ground up to fill the gap. Although not all samples could be measured by all devices,
thankfully some samples could be measured by several devices (both established
and unestablished). Data collected from these samples played a crucial role in
determining the uncertainty of the Bruker emission and Bruker reflectance
methodologies. When the Bruker emission and reflection methodologies resulted in
consistent resemblance of NREL FTIR and ET100 data for inter-compatible
samples, they could be considered valid.

As it would turn out, there were indeed initial limitations with the Bruker
emittance and Bruker reflectance methodologies which were uncovered after
noticing measurement discrepancies between the different methodologies. For
instance, originally the Bruker emission data would converge to the NREL FTIR data
only as measurement temperatures increased within the Bruker emission system
(e.g. measurements at 100 °C would disagree but measurements at 300 °C would
match NREL FTIR). This initial observation led to the eventual determination of
temperature variations as the primary source of uncertainty in Bruker emission
41

measurements. In the case of the Bruker reflectance system, it was quickly


determined that for smooth samples, the data closely matched the NREL FTIR data
set (as well as the ET-100) but for rough samples, there was very strong
disagreement.

The classic trial-and-error process of developing an experimental


methodology resulted in the development and refinement of two experimental
methods featured in this piece. Many of the limitations between the Bruker
reflectance and emittance methodologies were resolved or at the very least
identified and bounded during this process. Appendices B and C contain various
sets of superimposed data. Each figure compares data of the same sample,
collected through different methodologies.

2.3.1 Error in Bruker Reflectance

Bruker reflectance was limited by roughness and consistency. The spot size
was small enough to result in variations from surface heterogeneity and so several
measurements had to be taken across the sample, followed with an averaging of the
results. More importantly, this setup did not include an integrating sphere or other
mechanism for capturing diffuse reflection (unlike the ET100 and NREL FTIR).
Roughness could result in diffuse reflection which would invalidate the results from
this apparatus. Below a certain roughness, consensus was consistently achieved
between the Bruker Reflectance and the two established methods. However, beyond
a certain roughness, disagreement consistently occurred. The result for the roughest
“consensus” sample (RCS) measured using all three reflective radiometric
methodologies, is provided in Figure 2.16. This sample provided a limit on the
Bruker reflectance’s capability. Samples rougher than RCS were deemed
incompatible with Bruker reflectance while samples smoother than the RCS were
deemed compatible. For each sample that met the compatibility requirements, four
measurements were taken at different locations on the sample and then averaged
out to give final results. All comparisons between the Bruker Reflectance and NREL
42

FTIR measurements are featured in Appendix B. Here, Bruker Reflectance is plotted


alongside results of other methods when possible. Results are ordered in sequence
of sample roughness and so the divergence of the method as roughness increases
can be observed.

0.45
0.4 Bruker FTIR Reflectance
SOC FTIR Reflectance
0.35 ET100 Emissometer

0.3
0.25
0.2
0.15
0.1
0.05
0
0 5 10 15 20
Wavelength [ m]

Figure 2.16. 120 grit – finished SA508, the Roughest Consensus Sample. Good
agreement is achieved between the three reflectance methods. Every sample
measured by the three having a smoother surface than the RCS also achieved good
consensus while for samples that were rougher than RCS, Bruker reflectance data
diverged from the NREL FTIR data.

2.3.2 Error in Bruker Emittance

The validity and uncertainty of the direct emission measurement technique


used in this study is reviewed. Initially, it is important to verify the validity of the
method used in this study and to ensure that features such as atmospheric
absorption and background signal are adequately resolved. Once the method is
validated, uncertainty must be assigned to the measured emissivity values.
43

2.3.2.1 Validation of the Measurement Approach

We checked for various potential sources of error during this study. First,
oxidation may occur during measurement at elevated temperatures, compromising
the surface and invalidating the measurement. In this case, room-temperature
emissivity of a surface after an elevated-temperature measurement would differ from
the room-temperature emissivity of the surface before the elevated-temperature
measurement. Some examples of this hysteresis from oxidation can be found in
high-temperature measurements of stainless steel taken by Rolling et al.[75].
Second, absorption and emission from ambient CO2 and H2O may occur along the
measured wavelengths and affect the measured signal. Lastly, background emission
from the instrument, device surroundings, or some other unknown source could
have contributed to the received signal. Each of these sources of uncertainty are
discussed.

Oxidation during measurement was tracked by comparing the emissivity of


samples before and after elevated temperature measurement. Samples were always
measured at 100 °C before and after each 300 °C measurement and the two 100 °C
signals were compared. No sample produced significantly different spectra between
the two measurements as would be expected from additional oxidation (|ε1(λ,T) -
ε2(λ,T)| < .02 at low wavelengths where oxidation features would be expected to
occur).

Some approaches used to combat atmospheric absorption would include


taking measurements in vacuum or under argon or nitrogen purge. The last method
was used by Cao and others[2][15] to reduce atmospheric absorptive signal, but this
did not entirely resolve it. For this study, we measured the sample and the blackbody
in atmosphere within a few hours of each other under the same controlled
conditions. The reasoning behind this approach was that signal contribution from
atmosphere would “cancel out” in Eq. 2.3 as long as the atmosphere was similar
between the signal measurements of the sample and the blackbody reference.
44

,
,   , (2.3)
,

This approach required multiple measurements of the blackbody over the


course of the study and the measured signal of the blackbody reference could vary
considerably within the wavelength regions of atmospheric absorption between
different days. However, resultant emissivity measurements remained consistent
and reproducible.

Although the features from atmospheric absorption were not entirely resolved,
they were greatly mitigated by following this approach of simply taking
measurements of the sample and the blackbody reference under very similar
atmospheric conditions. Noise peaks associated with H2O (water moisture) were
mitigated to the extent that very small H2O features were only visible in the
emissivity spectra of nickel-based oxidative He-exposed alloys. However, noise from
CO2 remained in the spectra of most samples. Thankfully, this region is isolated to a
narrow range of 4.18 μm < λ < 4.4 μm and so the data from this region was
neglected and in the presented data, this region is merely represented as a line
connecting the data point at λ = 4.18μm to the point at λ = 4.4μm. The signal
surrounding 4.18μm < λ < 4.4μm is uncompromised and we assume that the values
within this corrected region behave similarly to the values just outside.

The method presented in the main text (referred to as the “simple method” for
the remainder of this supplementary section) neglects background. However, two
other additional methods incorporating background subtraction were used during this
study. The resultant emissivities calculated using the simplified method are
compared to results acquired using the background subtraction methods and the
differences appear negligible.

The first background subtraction method assumed background signal to be


temperature-dependent and sample-independent. This methodology, referred to as
45

the gold subtraction method, assumed instrument signal followed Eq. 2.4:

  , , , (2.4)

Where η is wavelength-dependent signal response of the instrument and SBG


is background signal. For this method, η was determined by measuring the
blackbody reference sample at 300 °C and assuming the signal was almost entirely
composed of the sample emission component of Eq. 2.4. If SR is measured and
εR(λ,T)E(λ,T) >> SBG(λ,T) where εR(λ,T) and E(λ,T) are known values, η may be
closely approximated by dividing SR by εR(λ,T)E(λ,T).

With Eq. 2.4 assumed, three objects are measured:

1) The sample of interest – ε is denoted without subscript


2) A very reflective gold film sample with εAu(λ,T) ideally near 0. In our case, 180
nm thick gold coating was vapor-deposited onto a fused silica wafer substrate
3) The emissive reference sample with a known emissivity, εR(λ,T)

Values of the gold reference sample are assumed to be similar to the literature
values reported by Aksyutov[76]. Each εAksytov(λ,T) is a basic fitting of Aksytov’s data,
which was conveniently measured at T = 300°C, the same temperature as the
measurements performed in this study. Two piecewise continuous linear functions of
wavelength were used in the fitting.

ε(λ,T) is calculated according to Eq. 2.5:

, ,
,   , (2.5)
, ,
46

Where S, SAu, and SR are the measured signals of the sample of interest, the gold
film sample, and the emissive reference sample, respectively.

The connection between the right side to the left can be shown by plugging in Eq.
2.4 into each S on the right side of Eq. 2.5. If we assume εAu(λ,T) ≈ εAkysutov(λ,T) and
that all temperatures and measured surface areas are equal, simple algebra results
in the right hand side of Eq. 2.5 collapsing into ε(λ,T) while eliminating all
background terms.

In Fig. 2.17, emissivities of two samples are calculated using both the simple method
and the gold subtraction method in order to compare the two methods. The samples
selected for presentation here exhibited more-substantial differences in emissivity
between the two methods compared to other samples. Nonetheless, for these
samples and for all other samples, the difference in emissivity between the two
methods is negligible.
47

( , T = 300°C)
N
Spectral Emissivity,

Figure 2.17 Spectral normal emissivities for Inconel 617 after 500-hour exposure to
1000 °C helium gas with oxidative impurities and reference Haynes 230 determined
using the simple method and the gold subtraction method. In the most extreme case,
emissivities differ by .04 between the two methods, but generally the difference is
.01 or less.

The second background-subtracting method, referred to as the self-subtraction


method, assumed the background to be sample-dependent and temperature-
independent, following 2.6:

  , , , (2.6)

Where α(λ) is some unknown environmental background function. This time,


cancellations fall through accordingly if ε(λ,T) is calculated according to Eq 2.7:

, ,
,   , (2.7)
, ,

Where, for this study, T1 = 300 °C and T2 = 100 °C. Four signal measurements are
taken per one set of emissivity data. The reference sample is measured once at
48

each of the two temperatures and the sample of interest is measured once at each
of the two temperatures. The key assumption in this method is that spectral
emissivity stays relatively constant between the measurements taken at 100 °C and
300 °C, although this is not necessarily always true since spectral emissivity can
indeed vary significantly over a difference of hundreds of degrees Celsius for many
materials. In Fig. 2.18, emissivities of two samples are calculated using both the
simple method and the self-subtraction method in order to compare the two
methods. The samples selected for presentation here exhibit greater difference in
emissivities calculated using the two methods compared to other samples.
Differences are negligible up until longer wavelengths between the two methods.
Here, the difference may be explainable by additional signal loss that occurs in the
self-subtraction method.
( , T = 300°C)
N
Spectral Emissivity,

Figure 2.18 Spectral normal emissivities for samples SC-B-2 and NA-A-2
determined using the simple method and the self-subtraction method. Outcome is
practically identical (< 1% relative difference) up until around 17.5 μm where
difference may be attributed to signal loss from the self-subtraction method.
49

Factors of background signal, atmospheric absorption/emission, and in-situ oxidation


have all been considered and resolved. The simple method is validated.

2.3.2.2 Measurement Uncertainty

The simple method assumes the following equation for the measurement of ε(λ,T):

, ,
,   , (2.8)
, ,

Uncertainty is assigned in the variables T, εR, and η(λ). Temperature may impact
spectral emissivity as well as the Planck Distribution associated with a given emitting
object. However, the potential deviations in temperature (between the blackbody
reference sample and the sample of interest) in this study are on the order of a few
degrees. While this change in a few degrees will noticeably impact the Planck
Distribution, a change of a few degrees should have no measureable impact on the
spectral emissivity of the objects being measured in this study and so spectral
emissivity is represented as temperature-independent.

2.3.2.2.1 Variation in Surface Temperature

The heated stage temperature is fixed in the experimental setup and so


samples of differing thermal resistance will also have differing surface temperature
under the same conditions. A thick, insulating sample will have a slightly lower
surface temperature than a thin, conducting sample for the same heated stage
temperature as illustrated in Fig. 2.19 (exaggerated). This is especially true if the
contacting surface area of Sample 1 is less than the contacting surface area of
Sample 2.
50

Stage Stage
Sample 2
Sample 1

Air Air

Tstage Tstage
T2

T1

Tamb Tamb

x x

Figure 2.19 Exaggerated temperature profiles of a thick insulating sample (left) and
a thin conducting sample (right) attached to the device stage. The raw measured
radiative power is a function of surface temperature (T1 and T2). Tamb represents
ambient temperature.
The potential difference in surface temperature is approximated from heat
transfer evaluations. Ultimately, the system is defined by sample thermal
conductivity (k), stage temperature (Tstage), sample thickness (x), and some heat
transfer coefficient from the sample surface to the surrounding air (h). We assume
the thermal resistance of the thermal paste to be negligible. While stage
temperature, thickness, and thermal conductivity are all fixed and known, the heat
transfer coefficient must be derived.

The system was treated as a representative vertical surface experiencing free


convection with room temperature air. The associated Prandtl, Nusselt and Rayleigh
numbers of the system were used along with equations developed by Churchill and
Chu[77] to determine h from convection. This value was added to a heat transfer
51

coefficient associated with radiative effects. The overall heat transfer coefficient,
including convection and radiation, is set to a generous value of 50 W/m2-K which is
slightly above the sum of the two calculated h’s (one for convection and one for
radiation) under maximal conditions. According to Eq. 2.7, error arises from surface
temperature differences between the sample of interest and the blackbody
reference. According to basic heat transfer calculations, a surface temperature
difference of 3.88 °C was determined for the blackbody reference and an Inconel
617 sample having exaggerated thickness (3.9 mm) and oxidation (50 μm of Cr2O3).
A ∆T of no more than 4 °C is expected to occur between the surface of any sample
and the surface of the blackbody reference sample.

According to Eq. 2.8, differences in surface temperature between the sample


and the blackbody reference result in uncertainty in calculated emissivity through
differences in the Planck distributions of the sample and the blackbody reference.
This difference is carried through by using the term F and ∆F, given by

,
  (2.9)
,

And

∆  | | (2.10)

Where T represents the surface temperature of the sample and TR gives the surface
temperature of the emissive reference sample. For a given ∆T (temperature
difference the sample and the blackbody reference), F is roughly the same around
the same general temperature range. For example, for ∆T = 4 °C

,  ° ,  °
,  ° ,  °
52

For this reason, F and ∆F are simply represented as functions of λ, T and ∆T,
where for this study, T is TStage and ∆T = 4°C which is a conservatively high bound
for the difference between the surface temperatures of the reference and any
sample. Curves of ∆F may be easily generated using Planck’s Distribution for
several different conditions. Examples of ∆F(λ, T = 300 °C, ∆T) are given in Fig.
2.20. ∆F(λ, T = 300 °C, ∆T= 4 °C) is the profile associated with this study and is the
center curve represented in red.

0.1
F( ,T = 300°C, T = 6°C)
F( ,T = 300°C, T = 5°C)
0.08 F( ,T = 300°C, T = 4°C)
F( ,T = 300°C, T = 3°C)
F( ,T = 300°C, T = 2°C)
0.06

0.04

0.02

0
4 6 8 10 12 14 16 18
Wavelength [ m]

Figure 2.20. F(λ, T = 300°C, ∆T). Error is limited as ∆T is limited. The largest region
of error occurs at shorter wavelengths where the temperature differential of the
Planck Distribution is greater.

2.3.2.2.2 Error in Reference Emissivity

Vertically-aligned carbon nanotube forests have proven to be capable of


absorbing light almost perfectly across a wide spectral range (0.2-200 μm) [78] and
53

so they make ideal candidates for laboratory blackbody references. Our blackbody
reference was a vertically-aligned carbon nanotube array from NanoTechLabs Inc.
Carbon nanotubes were grown on the silicon wafer perpendicular to the wafer’s
surface until the tubes reached a length of 100 μm. This reference had a uniform
spectral emissivity of .90 +/- .02 throughout the measured wavelength range.

The reference sample was calibrated by comparing results of the direct


emission measurement technique to results acquired through the reflectance
technique for samples of two well-known materials (sapphire and silica). First, the
reflectance technique was used to acquire an emissivity value for the sapphire
sample. Then, the signals of the sapphire and reference were measured and the
input value for εR was adjusted until good agreement of the resultant sample ε was
achieved between the two techniques. The emissivities acquired from these two
methods for the sapphire sample are provided in Fig. 2.21a. Using this new εR,
measurements of silica’s emissivity were compared between the two methods, this
time without adjusting εR. The emissivities acquired from the two methods for the
silica sample are provided in Figure 2.21b. The εR determined in this manner was
.90 +/- .02 across all measured wavelengths. This was the εR used throughout this
study.

(a) (b)
1 1
Spectral Emissivity, ( , ,T m)

Spectral Emissivity, ( , ,T m)

0.8 0.8
T m =75°C T m =75°C

0.6 T m =100°C 0.6 T m =100°C


T m =125°C T m =125°C
1-R Sapphire 1-R Silica
0.4 0.4

0.2 0.2

0 0
8 9 10 11 12 13 14 15 8 9 10 11 12 13 14 15
Wavelength [ m] Wavelength [ m]
54

Figure 2.21 Emissivity of sapphire (a) and silica (b) versus wavelength through two
methods of direct emission and reflection measurement. Calibration of emissive
reference is performed based on this measurement. εR = 0.90 +/- .02 is assumed in
direct emission measurement.
In every case, the sapphire and silica samples were sufficiently thick enough to
assume negligible transmission and so absorptivity should directly equate to
emissivity. Measurements were taken at several temperatures in the case of the
direct emission measurements to verify consistency. Measurements using the
reflectance technique were performed at room temperature.

2.3.2.2.3 Error in System Response

The system response, η(λ), depends on the quality of the signal being received. As
wavelengths approach zero or infinity, the Planck Distribution drops off in magnitude
and as a result, the signal received by our instrument is not adequate to overcome
noise and random variations in emission. This limited the range of our measurement
to 3 < λ < 18 μm. Within this range, the error is negligible and is not included in the
overall uncertainty.

2.3.2.2.4 Overall Uncertainty

Standard uncertainty evaluation was performed. If we assume the dominant


sources of error are (1) surface temperature differences between the blackbody and
the sample of interest, and (2) emissivity of the reference sample, the uncertainty is
straightforward. The uncertainty can be evaluated through partial derivatives,
yielding,


∆ , ,   ∆ , , (2.11)
55

The square root term is defined as the relative overall error factor, Q


  ∆ , , (2.12)

Using ∆εR/εR = .023 and the equation for ∆F(λ,T=300°C,∆T), Q is evaluated and
plotted in Fig. 2.22 for several different ∆T. Q(λ,T=300°C,∆T = 4°C) corresponds to
this study and is the center curve, plotted in red.

0.1
Q( ,T = 300°C, T = 6°C)
Q( ,T = 300°C, T = 5°C)
0.08 Q( ,T = 300°C, T = 4°C)
Q( ,T = 300°C, T = 3°C)
Q( ,T = 300°C, T = 2°C)
0.06

0.04

0.02

0
4 6 8 10 12 14 16 18
Wavelength [ m]

Figure 2.22 Relative Overall Error profile for direct emission measurements,
Q(λ,T=300°C, ∆T). F(λ,T,∆T) contributes more significantly to uncertainty at shorter
wavelengths. Q(λ,T=300°C,∆T = 4°C) corresponds to this study and is plotted in red.

The emissivity within uncertainty is given as ε(λ,θN) = εm(λ,θN) ± Q(λ,∆T) εm(λ,θN)


for each sample where εm is the measured emissivity.
56

3 Results and Analysis – RPV Materials (Low­alloy Steels) 

3.1 Spectral Directional Measurements of SA508

SA508 was initially exposed at 300 °C for 100 hours and 600 °C at 5 hours
to provide lower and upper boundaries (respectively) for the study. Results are
provided in Figure 3.1. The oxidation from 5 hours at 600 °C visibly dwarfed
oxidation from all other samples in this study, including the SA508 sample exposed
to 350 °C for 200 hours. It provides a good reference for an excessively oxidized
sample. Exposure temperatures are marked as Te in this study while temperature
during measurement is marked as Tm. The 600°C results match well to the work of
Cao[15] at 500°C and 700°C where both studies share the same wavelength range
of measurement. However, Cao focuses on a spectral range of 2 to 9 microns while
this piece focuses on a larger range from 1.7 to 25 microns.

At higher temperatures, the Planck Distribution shifts to favor shorter


wavelengths while at lower temperatures, longer wavelengths become more
significant. The measurements of Cao showed that very high emittance values
(roughly .8-.9) occur up to a wavelength of 9 microns for heavily oxidized SA508.
This study confirms the findings but shows that at longer wavelengths, spectral
emittance coefficients drop off sharply. This drop off should have minor
consequences for very high temperature environments, but greater consequences at
lower temperatures.
57

1
T e =600°C, 5 hr
T e =300°C, 100 hr

0.8 Reference
( , =10,T m=25°C)

0.6

0.4

0.2

0
0 5 10 15 20 25
Wavelength ( m)
Figure 3.1. Emittance measurements of SA508 air-oxidized at 300°C and 600°C.
Measurements were taken at room temperature.

3.1.1 Roughened SA508 measured at Room Temperature

Samples were also subjected to a few roughening conditions. Measurements


were taken at room temperature and results are given in Figure 3.2. The only
sample having a significantly different spectral emittance post-roughening would be
the shot-peened sample. According to the Profilometry measurements, the PV
(maximum peak-to-valley height) of the shot-peen surface was 6-21 times greater
than the other surface treatments. Although the condition exceeds the expected
surface roughness of an unoxidized reactor pressure vessel, it provides an upper
bound for the roughness effect. Assuming the roughness of an SA508 RPV stays
within the roughness scale of the samples roughened via the grinding papers,
roughness should not have a significant role in the surface emittance.
58

( , =10,T m=25°C)

Figure 3.2. Emittance measurements of roughened SA508.

3.1.2 Exposed SA508 measured at Room Temperature

Samples were subjected to oxidizing air at 350°C for several intervals before
also being measured at room temperature. Results are given in Figure 3.3. The
progression of the emittance coefficients matches very well with the experimental
measurements obtained by del Campo et al[14] for 99.8% pure iron. In particular,
the curve at 350°C, 200 hr matches well with Campo’s curves between 316 sec and
37 min at 480°C. Exactly how SA508’s progression might relate to bare iron’s would
require further study, most likely from a kinetics approach, but the resemblance of
the emissivities under differing oxidizing conditions makes sense since both
substrates should form similar oxide phases albeit at differing rates.

Overall, surface oxidation has significant influence in the emittance


coefficients of SA508 subjected to RPV conditions. Furthermore, in such conditions
59

it may take several hundred hours for total emittance to asymptote to a relatively
steady value.

1
Te =350°C, 200 hr
Te =350°C, 150 hr
0.8 Te =350°C, 100 hr
Te =350°C, 50 hr
( , =10,T m=25°C)

Te =350°C, 10 hr
0.6 Reference

0.4

0.2

0
0 5 10 15 20 25
Wavelength ( m)

Figure 3.3. Emittance measurements of SA508 air-oxidized at 350°C.

3.1.3 Unexposed SA508 measured at Elevated Temperatures

Additionally, some measurements of unoxidized mirror-finished samples


were taken at elevated temperature while under nitrogen purge in attempt to
investigate any temperature dependencies in SA508 spectral emittance (Figure 3.4).
Although the sample was under a nitrogen purge during the measurement, oxidation
still appears to have occurred at the higher temperatures. After measurements were
taken at 100°C, 200°C and 300°C, another measurement was taken at 100°C. The
discrepancy between the two measurements at 100°C implies that evolution of
spectral emissivity through these measurements can be attributed to oxidation
effects that occurred during measurement rather than optical shifts of the base
material. The effect of optical shifts (or lack thereof) could not be isolated.
60

0.5

0.45 Tm *=100°C
Tm =300°C
0.4
Tm =200°C
0.35 Tm =100°C
( , =10, Tm)

0.3

0.25

0.2

0.15

0.1

0.05

0
0 5 10 15 20 25
Wavelength ( m)

Figure 3.4. Reference SA508 measured at elevated temperatures. Tm* indicates the
second 100 °C measurement following high temperature characterization. The large
increase between the two measurements indicates oxidation occurred during
measurement.

Generally, metals exhibit spectral emittance having some kind of temperature


dependence[41]. This has been observed experimentally in the case of stainless
steel 304[75] and is observed for the case of T91 in this piece (discussed in
following sections). In both cases, a slight increase is observed across all measured
wavelengths. It is expected that a similar effect should occur for SA508. However,
the influence of temperature will be very small compared to the influence of oxidation
on the spectral emittance of SA508 in RPV conditions.
61

3.2 Spectral Directional Measurements of T91

Similar roughening conditions yield similar increases in emittance


coefficients between SA508 and T91. T91 exhibits much greater oxidation
resistance than SA508 due to increased chromium content.

3.2.1 Roughened T91 measured at Room Temperature

As with SA508, a few T91 samples were subjected to a few roughening


conditions. This time, only 3 conditions were evaluated since behavior should be the
same between the two alloys. Results are given in Figure 6. Once again, the shot-
peened sample features a significantly different spectral emittance profile; however,
the roughness features of this sample exceed that of a typical RPV surface.
Roughness is not expected to be a significant factor for the emittance of a T91
reactor pressure vessel surface.

0.4
Shotpeen
0.35 320 grit
Reference
0.3
( , =10,T m=25°C)

0.25

0.2

0.15

0.1

0.05

0
0 5 10 15 20 25
Wavelength ( m)

Figure 3.5. Emittance measurements of roughened T91.


62

3.2.2 Exposed T91 measured at Room Temperature

T91 was subjected to temperatures exceeding design-basis conditions for


RPV for a short duration. Nonetheless, the oxidation observed in the T91 sample
was very limited and the difference in spectral emittance between the exposed and
reference samples is relatively small excluding small wavelengths. Results are given
in Figure 3.6. The resultant spectral emittance curve appears to resemble that of the
SA508 exposed to 350°C for 10 hours and the curve of high-purity iron exposed to
air at 480°C for 8 seconds given by Campo. Although the temperatures and
durations are very different, the progression of the emissivity coefficient appears to
behave similarly to SA508 and bare iron.

This work yields relatively limited information on the long term spectral
emittance of T91 RPV. Based on the data from Figure 3.6, it is clear that T91 will
experience far less oxidation than SA508 and will have much smaller emittance. If a
conservatively-bounded value for the spectral coefficient is needed for the purposes
of heat transfer simulation, it would be safer to simply assume that a T91 RPV
maintains initial spectral emittance throughout operation. The real values are likely to
be higher but more work (using long-term exposure durations) would be required to
achieve a clearer picture.
63

1
Te =750°C, 5 hr
Reference
0.8
( , =10,T m=25°C)

0.6

0.4

0.2

0
0 5 10 15 20 25
Wavelength ( m)

Figure 3.6. Emittance measurements of T91 air-oxidized at 750°C for 5 hours.

3.2.3 Unexposed T91 measured at Elevated Temperatures

T91 has much greater oxidation resistance than SA508 and so elevated
temperature measurements could be pursued successfully. In Figure 8, the second
measurement at 100C converges to the first measurement taken at 100C,
demonstrating that additional oxidation played no observable role in the variation of
the emissivity coefficients taken during this set of elevated temperature
measurements. The variations in the spectra can be attributed to slight temperature-
dependent variations in the optical constants of the bare substrate material. This is
consistent with Rolling’s observations for stainless steel 304[75].
64

0.3
T m* = 100°C
T m = 300°C
T m = 200°C
T m = 100°C
0.2
( , =10,T m)

0.1

0
0 5 10 15 20 25
Wavelength ( m)

Figure 3.7. Emittance measurements of T91 at elevated temperatures.

3.3 Discussion – Low Alloy Steels

The uncertainty associated with the spectral directional measurement is


expected to be very small compared to the emission measurements. Due to the
accurate nature of the setup as well as the amount of data collected, calculations of
the total hemispherical emissivity coefficient could be performed.

3.3.1 Estimation of SA508 Total Emissivity from Data

It is common practice to input a single total hemispherical emittance


coefficient into heat transfer calculations. Total hemispherical emittance coefficients
are estimated based on the experimental curves obtained in the Figure 1.3 for 350C
oxidations. While the experimental data was obtained at a fixed angle near normal
65

incidence, normal emissivity can roughly represent total emissivity under certain
circumstances. In the case of metals, hemispherical emissivity tends to exceed
normal emissivity while in the case of dielectrics, hemispherical emissivity tends to
behave similarly to normal emissivity (.94<ε/εN <1.03 while 1>ň>6)[41]. For
conservative heat transfer simulations, we suggest that an emissivity coefficient
could be estimated by the total emissivity coefficient at 10 degrees incidence which
is based on the integration of Figure 3.2 or 3.3, following equation 3.1.

∫ ε λ (T ) Eλ (λ ,T )dλ
ε (T ) = λ =0
σT 4 (3.1)
It is not possible to measure individual spectral values across all wavelengths
of light. The uncertainty based on the incompleteness of any experimental
measurement in spectra can be found by integrating the Planck Distribution over the
unmeasured wavelength ranges for the bounding cases. At higher temperatures,
radiative power concentrates at shorter wavelengths but is more evenly distributed
at lower temperatures. Using the Planck Distribution, at temperatures of 500°C,
approximately 79% of radiated power occurs between 2 and 9 microns while 98%
occurs between 1.7 and 25 microns. As stated previously, LWR reactor pressure
vessels can operate at temperatures as low as 260°C. At 260°C, approximately 61%
of radiated power occurs between 2 and 9 microns while 96% of radiated power
occurs between 1.7 and 25 microns.

With this in mind, the total emittances of the SA508 samples exposed to
350°C air are calculated through numerical integration of the directional
measurements within the wavelength bounds of measurement. This gives an
approximation at the actual value.

, ,
  (3.2)
,
66

Next, the lower uncertainty bound is determined. The minimum that the value could
possibly be would be if the rest of the unmeasured spectral emissivity equaled 0,
according to Equation 3.3.

, , ,   , , ,
   (3.3)
,

The upper uncertainty bound is determined. The maximum that the value could
possibly be would be if the rest of the unmeasured spectral emissivity equaled 1,
which is expressed in Equation 3.4

, , ,   ,
  (3.4)
,

Using the acquired experimental data, this evaluation is performed for four
operational temperatures: 260, 290, 320 and 350°C. Optical shifts are neglected and
Planck shifts are maintained. The center value is based on Equation 3.2 while the
lower and upper bounds of the error bars are determined by Equation 3.3 and
Equation 3.4, respectively. Figure 9 uses the data from the samples oxidized at
350°C. Practically, this figure answers the question: Neglecting optical shifts, what
would ε(θ=10,T) of the experimentally oxidized surfaces be at different
temperatures? For clarity, the error featured in the figures 9-14 illustrates the error
from the evaluation method, and does not represent overall error.
67

0.8
Te =350°C, 200 hr
0.6 Te =350°C, 150 hr
Te =350°C, 100 hr
Te =350°C, 50 hr
0.4 Te =350°C, 10 hr
Reference

0.2

0
260 270 280 290 300 310 320 330 340 350
Temperature (°C)

Figure 3.8. Estimated total angular emittance coefficients of SA508. Values are
calculated using equations 1-4 on data from samples oxidized at 350°C for various
exposure times.

The progression of the total calculated emittance across time is given in figure 3.9
for T=350°C. Large increases occur initially before approaching a limit that likely lies
around .8 (based on the 600°C exposure data).
68

0.8
( =10,T=350°C)

0.6

0.4

0.2

0
0 50 100 150 200
Exposure Time (hr)

Figure 3.9. Calculated ε(θ=10,T=350°C) of SA508 samples through various points of


air-exposure.

The evaluation is also performed using the 600°C, 5 hour results and then on
the data gathered by Cao et al. at 500°C for 4 hours and 700°C for 5 hours.
Temperatures of 350°C, 400°C, 500°C and 600°C are used. Each sample
represents a heavily oxidized SA508 surface whose emittance has likely reached a
steady state. The data gathered in this study is used in Figure 3.10 while the data
gathered by Cao is used in figure 3.11.
69

Figure 3.10. Estimated total angular emittance coefficients of SA508. Values are
calculated using equations 3.1-3.4 on the data collected for the sample oxidized at
600°C for five hours. Experimental data was acquired at room temperature.

Figure 3.11. Estimated total normal emittance coefficients of SA508. Values are
calculated using equations 3.1-3.4 on the data collected by Cao for the sample
oxidized at 500°C for four hours (a) and for the sample oxidized at 700°C for five
hours (b). Experimental data was acquired in-situ at 500°C and in-situ at 700°C,
respectively.

There are two possible reasons for the minor difference between our results
and Cao’s. The first immediate reason would be difference in spectral range
between the two experiments. Our study includes longer wavelengths where
emissivity coefficients are significantly smaller (see Figure 3.1). Secondly, emittance
70

was measured directly at temperatures of 500°C and 700°C in the work of Cao while
our measurements were taken at room temperature. In other words, Cao’s
measurements directly account for optical shifts. Although this study could not
isolate the temperature dependency in SA508, some dependency was observed in
T91 and may occur in SA508 at high temperature. In both studies, the range of 2 to
9 μm was included. Within this range, our emissivity data is slightly less than the
results of Cao, on average. Although the work of Cao was measured at normal
incidence while this study was measured at 10° incidence, this should have little to
no significance. Either way, the total emissivity values calculated from this study lie
well within the lower bounds from the values calculated using Cao’s results
according to Equations 3.1-3.4.

3.3.2 Estimation of T-91 Total Emissivity from Data

Three different criteria were evaluated for T91. In the case of roughness,
significant differences in emittance coefficients were observed only after samples
attained a roughness value greater than that which would be expected of an
operating reactor pressure vessel. In the case of oxidation, T91 was exposed to
conditions exceeding that of design-basis transient scenarios for Gen IV reactor
pressure vessels and an increase in coefficients was observed at lower
wavelengths. In the case of temperature, a small increase in temperature across
wavelengths was observed.

Based on the data, only small increases were observed in T91 after 750°C
exposure and we speculate that the emittance coefficients for a reactor pressure
vessel composed of T91 through steady state operation will not experience the large
increases observed with SA508. For heat transfer simulations of systems having a
T91 reactor pressure vessel, a conservatively low value can be obtained through the
integration of the curves of Figure 3.6 or 3.7. Since there was very little change
observed in coefficient during this study, it would be safe to simply assume an
unchanged surface and coefficient from the unexposed sample. The integrated
71

exposed and unexposed results, neglecting optical shifts, are presented in Figure
3.12.

( =10,T)

Figure 3.12. Estimated total angular emittance coefficients of T91. Values are
calculated using equations 1-4 on the data collected for each sample used. The
exposed sample was subjected to 750°C air for 5 hours before measured at room
temperature. The bare samples were unexposed mirror-finished samples measured
while at 300°C and 25°C, respectively.

As temperatures vary, the Planck distribution will also adjust and so the
relative contribution of emission within a given spectral range to total emission will
also vary. At higher temperatures, spectral coefficients at shorter wavelengths carry
greater relative significance than what they would at lower temperatures due to the
Planck shifts. Based on the work of Campo, as well as the work shown in this study,
minor oxidation in iron or iron-based alloys greatly increases spectral emittance at
short wavelengths while leaving spectral emittance at larger wavelengths relatively
unchanged. This pattern is visible in Figure 3.6, where initial oxidation yields a
72

dramatic increase in emittance at wavelengths below 5µm, but emittance at longer


wavelengths remains largely unchanged.

The preferential shape of spectral emittance results in a total emittance that


varies across temperature. In the case of unoxidized steels, the tendency of the
spectral emittance to be greater at shorter wavelengths results in a total emittance
coefficient that increases with temperature. As temperatures rise, the Planck
Distribution will shift to favor shorter wavelengths, where higher spectral emittance
coefficients are located. The preferential profiles in Figure 3.6 lead to the
temperature-dependent trends demonstrated in Figure 13. Unoxidized steel
maintains a somewhat preferential shape while lowly-oxidized low-alloy steel or
lowly-oxidized iron maintains a very preferential shape. Since T91 will likely
experience little oxidation in reactor environments, the temperature dependence
arising from Planck shifts will be noticeable, as is the case in Figure 13.

Optical shifts have a small impact on the total emittance coefficient of T91.
This impact can be neglected for shorthand approximation but if a greater sense of
accuracy is desired, two potential options could be pursued. Optical shift accounting
can be accomplished through further measurement in-situ at elevated temperatures
or potentially through the application of physical models. Drude Theory has found
some success predicting values for stainless steel 304.[32] Other methods of
deriving optical constants for metals exist and a decent review of some can be found
in the thesis of Teordorescu[79].
73

4 Results and Analysis – Structural Alloys 
There are several mechanisms through which exposure to various high-
temperature environments can modify the emissivity of alloys. A change in surface
chemical composition from exposure can result in a large change in the optical
constants of the surface, thereby changing the emissivity. If the chemical change in
the surface results in a formation of a thin layer, thin-film interference can
significantly affect the spectrum [80][81]. For films of sufficient thickness and optical
absorption, the emissivity will simply be the bulk emissivity of the top layer. Finally,
exposure may result in rougher surfaces with greater surface area, and roughening
generally increases emissivity [10][43]. A ZYGO white light interferometer was used
to measure the RMS roughness of each sample.

Within the measured emissivity spectra, some features appeared due to


atmospheric absorption. The peak associated with CO2 absorption (λ ~ 4.2 μm) [82]
was fitted and subtracted. Features in the 5-8 μm range are due to atmospheric
absorption by H2O [82]. These features were prominent in the data presented for
oxidizing-helium-exposed Haynes 230, oxidizing-helium-exposed Inconel 617, and
sCO2-exposed stainless steel.

4.1 Helium Gas Environment

The H2O addition to the helium gas yielded a strongly oxidative environment to
both nickel-based alloys. Both Inconel 617 and Haynes 230 formed thick, black
oxide layers in this environment (according to backscattered electron micrograph
analysis conducted on samples exposed to similar conditions [83]), and thus
exhibited a dramatic increase in emissivity (Fig. 2). For both oxidizing-helium-
exposed nickel-based alloys, as well as sCO2-exposed Haynes 230 presented in
section 5.4, material resonances occur between 16 < λ < 18 µm. The surface oxide
layer of these samples is expected to be predominantly Cr2O3 [83], which exhibits
two high-intensity absorption bands roughly between 15 and 18 µm [82].
74

In the carburizing environment, reactions of Cr with CO or CH4 produce


chromium carbide (assumed to be Cr23C6 [84]. In addition to changing the
composition of the surface, the carburizing environment yielded large increases in
roughness. Overall, a moderate increase in emissivity is observed for both the
Haynes 230 and the Inconel 617 alloys. The measurements of the Haynes 230
samples are provided in Fig. 2 while the Inconel samples are given in Fig. 3.
Photographs of the samples are provided next to the spectral emissivity profiles. The
RMS roughness values of the helium-exposed samples are provided in Table 4.

1
0.9
0.8
(a)
0.7
0.6
(a) Ox. Env.
0.5 (b) Carb. Env.
(c) Ref. (b)
0.4
0.3
0.2 (c)
0.1
0
4 6 8 10 12 14 16 18
Wavelength [ m]

Figure 4.1. Spectral normal emissivity for Haynes 230 after 500-hour exposure to
1000 °C helium gas. Measurements taken at 300 °C. Photos of the samples are
provided to the right of the spectra.
75

1
0.9
0.8
(a)
0.7
0.6
(a) Ox. Env.
0.5 (b) Carb. Env.
(c) Ref. (b)
0.4
0.3
0.2 (c)
0.1
0
4 6 8 10 12 14 16 18
Wavelength [ m]

Figure 4.2. Spectral normal emissivity for Inconel 617 after 500-hour exposure to
1000 °C helium gas. Measurements taken at 300 °C. Photos of the samples are
provided to the right of the spectra.

4.2 Molten Salt Environment

The most significant corrosion mechanism in structural alloys in molten fluoride


salts is chromium depletion. Therefore, Hastelloy-N, which was specifically designed
for molten fluoride salt applications, contains only about 6-7% Cr. Thus, the surface
experiences relatively low corrosion and the emissivity coefficients remain relatively
unchanged as a result of FLiNaK exposure (Figure 4.3).
76

1
FLiNaK Env.

0.8

0.6 Ref.

0.4

0.2

0
4 6 8 10 12 14 16 18
Wavelength [ m]

Figure 4.3 Spectral normal emissivity for Hastelloy N exposed to 850°C FLiNaK for
1000 hours. Measurements taken at 300 °C. Photos of the samples are provided on
the figure (background cropped out).
316 stainless steel contains greater concentration of chromium (~17%) than
Hastelloy N, so chromium depletion for this material occurs more readily. A good
overview of the process of chromium depletion in 316 stainless steel is provided in
an ORNL report by Keiser [41].

The reflectance technique was used for the FLiBe-exposed 316 stainless
steel (Fig. 4.4) and the emissivity of 316 stainless steel experiences a significant
increase after exposure. To account for localized surface heterogeneities, four
measurements were taken and an average of the measurements is presented. The
standard deviation associated with the spectral measurements is provided by the
shaded region in Fig. 4.4.
77

FLiBe Env.

Ref.

Figure 4.4 Spectral directional emissivity for SS-316 exposed to 750 °C LiF-BeF2 for
1000 hours. Measurements were taken at room temperature using the reflective
radiometric technique, with a numerical aperture of 0.4.

4.3 Liquid Sodium Environment

Although decarburization of alloys in long-term exposures in molten sodium


has been reported, corrosion rate of nuclear alloys in liquid sodium is generally
considered to be quite low compared to other reactor-relevant environments [42].
Consistent with this, the emissivity of the exposed 316 stainless steel samples
remained relatively unchanged after sodium exposure (Fig. 4.5).
78

1
Sodium Env.

0.8

0.6 Ref.

0.4

0.2

0
4 6 8 10 12 14 16 18
Wavelength [ m]

Figure 4.5 Spectral directional emissivity for 316 stainless steel after 1000-hour
exposure liquid sodium at 650 °C, Measurements taken at 300 °C. Photos of the
samples are provided on the figure (background cropped out).

4.4 sCO2 Environment

Figure 4.6 shows the spectral emissivity measurements for Ni-based alloy
Haynes 230 after exposure to sCO2 at 650 °C. The surface oxide layer of these
samples is expected to be predominantly Cr2O3 and small amounts of Cr2MnO4
based on previous exposure studies [21]. The spectral emissivity increases, but
does not reach as high a value as the oxidizing-He-exposed Haynes 230. Similar to
the oxidizing helium exposure case, a peak in the emission data is observed for the
Ni-based alloy between 16 < λ < 18 µm. This is possibly associated with the
absorption bands of Cr2O3, which occur roughly near that region [36][43].
79

sCO2 Env.
0.8

0.6
Ref.

0.4

0.2

0
4 6 8 10 12 14 16 18
Wavelength [ m]

Figure 4.6. Spectral normal emissivity for 650°C supercritical CO2 – exposed
Haynes 230. Measurements taken at 300 °C. Refer to Table 2.4 for sample info.

As shown in Fig. 4.7, 316 stainless steel experiences a much greater increase
in emissivity than Haynes 230 under the same test conditions. As would be expected
from previous literature [44], the iron-based 316 stainless steel experienced greater
oxidation than the nickel-based Haynes 230. Olivares [45] performed high-
temperature sCO2 exposure studies of 316 stainless steel under similar conditions to
those in this study and reported that iron-rich oxide layers dominated the topmost
surface layer through all exposure durations (including 500 hours) [45]. Other work
also report multilayer oxide formation with iron-rich oxides on the outermost layer
under related conditions [46].

Our surface grew a rough, grey, and visibly thick surface layer which appeared
very similar to the surfaces of oxidized low-alloy steels (having grey, iron-rich oxides
on the surface) presented in previous work [47]. We assume the top oxide layer is
iron-rich. The exposed 316 stainless steel exhibited an RMS roughness of 1.624 +/-
0.535 µm while the reference exhibited .0263 +/- 7.024E-3 µm.
80

1
sCO2 Env.

0.8

0.6
Ref.

0.4

0.2

0
4 6 8 10 12 14 16 18
Wavelength [ m]

Figure 4.7 Spectral normal emissivity for stainless steel 316 after 400-hour exposure
to 650 °C supercritical CO2. Measurements taken at 300 °C. Photos of the samples
are provided on the figure (background cropped out).

No material resonances occur in the spectra of the sCO2-exposed 316


stainless steel up to λ = 18 µm. If we assume iron-rich oxides form on the surface of
this sample, the topmost oxides should either be magnetite (Fe3O4) [46] or possibly
a very thin layer of hematite (Fe2O3) on top of magnetite (Fe3O4) [45]. Fe2O3 exhibits
two absorption peaks very similar to those of Cr2O3 mentioned previously, however,
they are shifted to longer wavelengths by 2-5 µm [36]. If we were to expect a
material resonance in the sCO2-exposed 316 stainless steel associated with Fe2O3,
analogous to the material resonance in the sCO2-exposed Haynes 230 (λ ~ 17 µm)
assumed to be associated with Cr2O3, it would likely be past the upper measurement
limit of 18 µm. Fe3O4 also exhibits absorption bands in the infrared. These occur at
wavelength ranges between the wavelength ranges of Fe2O3 and Cr2O3 absorption
bands [48]. While the absorption bands of Fe3O4 are rather distinct below 120 °K, the
absorption bands are much more muted at room temperature and above [49][50]
since the Verwey transition has been passed [51]. In our previous work [47] [52],
81

material resonances of varying magnitude occur in the spectral emissivity of oxidized


low-alloy SA508 steel around the wavelength ranges of 17.5 < λ < 24 µm. These
SA508 alloys form iron-rich oxides on the surface and the material resonances
observed in their spectra may be associated with absorption bands of iron oxides
which occur around these wavelength regions. We postulate that material
resonances may occur in the spectra of sCO2-exposed stainless steel at
wavelengths just above the upper measurement limit of this study.

5 Modeling  
Emissivity is a thermodynamic property of a material (surface), but its first
principle computation is difficult. Here, we use a model that utilizes Kirchoff’s law
relating spectral and directional emissivity to reflectance, and optical constants of the
material. In short, the model progresses as follows: A wavelength is selected by
sampling the Planck Distribution. Then, many light rays at this wavelength are
simulated against an input surface at various different angles of incidence and at
various different locations along the surface profile. Spectral directional emittance
can be related to spectral directional reflectance through Kirchhoff’s Law and so this
produces a set of spectral directional emittances. This set can then be integrated to
obtain total spectral emittance of the surface at this wavelength. A new wavelength
is sampled and the procedure repeats to obtain total spectral emittance at the new
wavelength. Once the Planck Distribution has been sampled sufficiently, the total
spectral emittance values may be averaged to yield a total hemispherical emittance.

5.1 Theory

Emissivity is a thermodynamic property of a material (surface), but its first


principle computation is difficult. Here, we use a model that utilizes Kirchoff’s law
relating spectral and directional emissivity to reflectance, and optical constants of the
material. In short, the model progresses as follows: A wavelength is selected by
82

sampling the Planck Distribution. Then, many light rays at this wavelength are
simulated against an input surface at various different angles of incidence and at
various different locations along the surface profile. Spectral directional emittance
can be related to spectral directional reflectance through Kirchhoff’s Law and so this
produces a set of spectral directional emittances. This set can then be integrated to
obtain total spectral emittance of the surface at this wavelength. A new wavelength
is sampled and the procedure repeats to obtain total spectral emittance at a the new
wavelength. Once the Planck Distribution has been sampled sufficiently, the total
spectral emittance values may be averaged to yield a total hemispherical emittance.

The evaluation method of total spectral emittance used in the GOA follows
Figure 5.1. This model differs slightly from several other GOA adaptations in that
BRDF’s are not used and spectral directional values are pursued directly.

L
Figure 5.1. The progression of the GOA towards a spectral emittance. The first step
(1) involves projecting a ray of light onto the input surface and evaluating for the
reflected and absorbed intensity. The Fresnel Equations at the particular spot on the
profile are used for this evaluation, as explained in section 2.1. The second step (2)
involves starting the incident ray at a new location on the plane of incidence and
performing the same progression. The average of satisfactorily many evaluated
starting locations, uniformly distributed across the incident plane, yields the
directional emittance for that surface as explained in section 2.2. A new incident
angle is evaluated in the third step (3). At this new incident angle, a new spectral
directional emittance is evaluated using steps (1) and (2). An adequate number of
directional emittances may be integrated to yield a total spectral emittance as
explained in section 2.3.
83

5.1.1 Evaluation of ε(s,λ,θ,T) at a surface point s

Along individual points, spectral directional emittance are calculated. In accordance


with Kirchhoff’s Law, it follows that

, , ,     , , , (5.1)
It is also assumed that Kirchhoff’s Law is valid for the two different polarizations of
light.

, , ,   (5.2)

(5.3)

, , ,   (5.4)

The reflection coefficients for each polarization are found by

  (5.5)

  (5.6)

Where R 1 and R 1 represent the relative intensities of the incident s-polarized and
p-polarized light, respectively. Meanwhile, R 2 and R 2 represent the corresponding
relative intensities of the reflected s-polarized and p-polarized light. The new ray
could in theory reflect again on the surface somewhere. If this occurs, new values of
R 2 and R 2 are simply calculated using the most recently calculated R 2 and R 2

input as R 1 and R 1 into equations 5 and 6. The very first R 1 and R 1 (original
incident values) are set to 1. The R 2 and R 2 coefficients found for the final ray to
strike the surface are input into Equation 5.4 for R and R .

Terms rs and rp may be calculated according to Equations 5.7 and 5.8

  (5.7)

  (5.8)
 

Where coefficients A and B are


84

1
      4        
2

1
      4        
2

Algebraically, temperature and wavelength dependencies of ε(λ,θ,T) originate from


the temperature and wavelength dependency of n and k. However, oftentimes the
temperature dependency of the optical constants may be neglected. The angular
dependency of ε(λ,θ,T) originates from Equations 5.7 and 5.8.

5.1.2 Evaluation of ε(λ,θ,T) across the surface

The spectral directional emissivity, ε(λ,θ,T), of the surface is determined by


the emissivity of each point on the surface and so

, ,   , , , (5.9a)

In the model, equation 5.9a is approximated by 5.9b over a 1D surface.

, ,   , , , ∑ , , , (5.9b)

Where N ≥ 100

In other words, sufficiently many spot measurements averages out to the value of
the surface.

5.1.3 Evaluation of ε(λ,T)

Integrating ε(λ,θ,T) over all solid angle gives ε(λ,T) according to Equation 5.10.

  , , , (5.10)

If ελ(θ,φ,T) is equivalent across the azimuthal angle, (5.10) can be simplified to:
85

  , , (5.11a)

In the model, this equation is approximated by

∑ , , sin  cos  (5.11b)

Where

 
2 1

And generally,

N = 89

Pseudorandom random sampling of the incident angle may be more desirable in the
future.

5.1.4 Evaluation of ε(T)

It is well known that total hemispherical emittance can be gathered from total
spectral emittance via equation 5.12a.

,
  (5.12a)

This work resolves the wavelength portion of the Equation 5.12 by running a
spectral GOA for a series of wavelengths that are representative of the Planck
distribution and then averaging the outputs. A representative sample of wavelengths,
λj, is generated through Monte Carlo sampling of the cumulative distribution function
for energy as a function of wavelength. The probability density function used is
proportional to the Planck Distribution for some temperature. For most dielectrics
and metals, the most dominant effect that temperature has on ε(T) manifests from
the inherent temperature dependence of the Planck Distribution. For the remainder
of this piece, shifts in temperature that shift the Planck Distribution will be referred to
as Planck shifts. So the total hemispherical emissivity is given as
86

  lim   (5.12b)

And generally,

M ≥ 100

Where the λ’s are generated through Monte Carlo sampling of the Planck
Distribution. M depends on the degree of certainty that is desired and the variance of
the emissivity values across wavelength. In this study, all generated total
hemispherical emissivity values are the average of four runs of M=100 each. The
error bars represent the standard deviation amongst these calculated total values.
The cumulative distribution function of the wavelength (λ) is given in the right hand
side of Equation 5.13a.

,
  (5.13a)
,

Where pj is a randomly generated number between 0 and 1. Equation 5.13a can


rearrange to:

,   , (5.13b)

λi can now be solved for numerically using Equation 5.13. Higher temperatures
result in a higher relative frequency of shorter wavelengths and vice-versa. The
Monte Carlo sampling used in this work effectively imitates this dominant effect. Now
the λ’s generated in this manner are used in Equations. 5.1 through 5.12b.

5.2 Model Results

The modeling of compositional, temperature, and roughness dependencies is


approached iteratively. The final model, which attempts to demonstrate how each
dependency can be accounted for in a total hemispherical emissivity coefficient
calculation, includes a few deliberate steps of adjustment and extension. However,
the model is built progressively to enable validation for each step individually. Each
87

case is used to calculate normal spectral emissivities followed by spectral directional


emissivities and then finally total hemispherical emissivities. In each case,
comparisons are made with experimental literature where possible and total
hemispherical emissivities always account for Planck shifts. Conveniently, the GOA
method can still be applied to a flat surface and so comparison and validation can be
achieved prior to the modeling of rough surfaces. For flat surfaces, it is relatively
easy to produce an analytical or numerical solution of the surface since any given
point along the flat surface is representative of the surface as a whole.

The model is started as GOA of a flat platinum surface so that it could be


compared to existing and reliable literature. Platinum is a relatively chemical inert
material and so unintended effects of chemical reactions that may have occurred in
experimental studies ought to be limited. In this case, dependency on composition,
and consequently, dependency on associated optical constants. Here, the optical
constants are taken from the Rakic [40] where optical constants were generated by
fitting the data acquired by Weaver[85] using the Lorentz-Drude model. Temperature
dependence of the optical constants is neglected since Silvera has suggested that
the optical constants of platinum within the infrared wavelength range remain
relatively constant across temperature.[86]

The second case of the GOA introduces Drude theory for materials of
stainless steel and tungsten in the attempt to capture temperature dependency in
optical constants in addition to composition dependency and Planck shifts. Here, the
optical constants are not assumed constant across temperature. The results are
compared to the work of Rosenberg et al. [32] whose analytical model also uses
Drude theory to derive temperature-dependent model inputs in the same manner for
a flat surface.

The third case of the GOA attempts to mimic effects of roughness in addition
to composition dependency. It is well established that the GOA method loses validity
for directional spectral values at higher incident angles. It is shown in the spectral
directional output during the first and second cases that p-polarized light is greatly
88

overestimated as incidence approaches 90 degrees. A tan(θ) term is introduced to


roughly represent the high angle scattering effects that are typically neglected with
the GOA method. The pseudo-random Fourier series surface profile is used and,
once again, platinum is used as a benchmark and secondary temperature
dependencies of the optical constants are neglected. The same references used in
the first case are used again for optical constant input.

The final case of the GOA combines the optical temperature dependency
accommodations of the second case with the roughness dependency
accommodations of the third case. Drude theory is once again used to derive the
optical constants and pseudorandom Fourier series profiles are used to characterize
the surface. The tan(θ) term is used to account for high angle scattering. Surfaces of
stainless steel 304 and tungsten are modeled.

Table 5.1. Details of Presented Model Evaluations for each if the four Cases
n, k Dependencies and Effects
Surface Emittances
Case Derivation Accounted for during
Composition Evaluated
Method Evaluation
εN(λ) Composition, λ, Polarization
ε||(θ,λ)/εN(λ),
1 Pt Experiment Composition, λ, Polarization, θ
ε (θ,λ)/εN(λ)
Composition, λ, Polarization, θ,
ε(T)
E(λ,T)
ε|| (θ,λ,T)/εN(λ,T),
Composition, λ, Polarization, θ, T
ss304, Drude ε (θ,λ,T)/εN(λ,T)
2
W Theory Composition, λ, Polarization, θ, T,
ε(T)
E(λ,T)
Composition, λ, Polarization,
εN(λ)
Roughness
ε||(θ,λ)/εN(λ), Composition, λ, Polarization, θ,
3 Pt Experiment
ε (θ,λ)/εN(λ) Roughness
Composition, λ, Polarization, θ,
ε(T)
E(λ,T), Roughness
ε||(θ,λ,T)/εN(λ,T), Composition, λ, Polarization, θ, T,
ε (θ,λ,T)/εN(λ,T) Roughness
ss304, Drude Composition, λ, Polarization, θ, T,
4 ε(θ,T)
W Theory E(λ,T), Roughness
Composition, λ, Polarization, θ, T,
ε(T)
E(λ,T), Roughness
89

5.2.1 Case 1: Flat Platinum Surface

Optical Constants are taken from Rakic’s fitting of the data compiled by
Weaver. These experimental measurements were taken at temperatures either
unreported, or significantly below that of Rolling and Silvera (250 C in the case of the
work of Seignac[87]).
First, normal spectral emissivity coefficients are calculated and compared to
experimental literature (Figure 5.2). Model output aligns well with the experimental
work of Silvera[86] and Aksyutov[76]. Consistency between the data points of
Silvera, Aksyutov and the model supports the notion of temperature-independent
optical constants of platinum. Slight variation is found in Rolling’s[75] work at shorter
wavelengths, however, rougher samples in the work of Rolling tended to have
greater values at lower wavelengths.

Figure 5.2. Normal Spectral Emissivity Coefficients of Platinum. The model assumes
temperature independence. Good agreement is achieved between the model and
the works of Silvera and Rolling. Model values tend to be slightly less than
experimental values.
90

Relative spectral directional emissivity coefficients are evaluated and


compared to the experimental work of Rolling. Figure 5.3 shows the evaluation of
the parallel-polarized component of light across angle for several wavelengths while
Figure 5.4 shows the same for the perpendicular-polarized component of light. The
parallel component shows large increases at longer wavelengths and has maximum
peak in between 80 to 90o while the perpendicular component remains unchanged
across wavelengths and decreases as the angle increases. Both of these tendencies
are visible in the experimental literature as well as in the model output. However, the
model tends to slightly overestimate the intensity of emitted parallel-polarized light.
As mentioned previously, the GOA method does not perfectly resolve scattering at
high angles of incidence.

5
=6 m
4.5 =4 m
=3 m
4 =2 m
=1.5 m
3.5
( )
( )

3
N

( , )/
( , )/

2.5
||
||

1.5

0.5
0 10 20 30 40 50 60 70 80 90
(deg)

Figure 5.3. Parallel Component of Platinum Relative Directional Spectral Emissivity


Coefficients through various wavelengths. Rolling experimental (left) and model
output (right) both show increases at longer wavelengths, particularly at higher
angles of incidence.
91
0.5

0.4

( )
0.3 =6 m

N
=4 m

( , )/
=3 m
=2 m
0.2 =1.5 m

0.1

0
0 10 20 30 40 50 60 70 80 90
(deg)

Figure 5.4. Perpendicular Component of Platinum Relative Directional Spectral


Emissivity Coefficients. Rolling experimental (left) and model output (right) both
show little to no change across wavelength.

The model is Monte Carlo – sampled according to the Planck Distribution to


generate total hemispherical emissivity coefficients across temperature. This
comparison is given in Figure 5.5. Although variation can occur between
experimental results, the current model well predicts the tendency of experimental
results.
92

0.3
Model
Sully
Davisson
Foote
Goard
0.2 Jain
Rolling
(T)

0.1

0
500 700 900 1100 1300 1500
Temperature (K)
Figure 5.5. Model and Experimental values [88][89][90][24][91][75]for total
hemispherical emissivity of platinum.

5.2.2 Case 2: Flat 304 SS Surface

The model is used once again for a flat, uniform surface profile and compared
to experimental relative spectral directional emissivity coefficients for both
polarizations. Rather than use literature values of optical constants, the optical
constants are now derived using Drude theory as is done in Rosenberg et al. [3].
The benefit to Drude is that now the temperature dependence of the optical
constants is accounted. The presented model is compared to Rosenberg’s work to
demonstrate convergence to a closed-form method in the case of a flat surface.

The relative spectral directional emissivity coefficients are once again


evaluated and compared to the experimental work of Rolling. Figure 5.6 shows the
evaluation of the parallel polarized component of light across angle for several
wavelengths while Figure 5.7 shows the same for the perpendicular polarized
component of light. Both of the tendencies observed in the experiments are visible in
93

the model output. However, the model tends to still overestimate the intensity of
emitted parallel-polarized light at high angles of incidence as well as predict a
greater angle of peak parallel emittance.

3
=6 m
=3 m
2.5 =1.5 m

2
( ,T)
N
( , ,T)/

1.5
||

0.5

0
0 10 20 30 40 50 60 70 80 90
(deg)

Figure 5.6. Parallel Component of 304 Stainless Steel Relative Directional Spectral
Emissivity Coefficients. Rolling experimental (left) and model output (right) both
show increases at longer wavelengths, particularly at higher angles of incidence. T =
953 °K.
94
( ,T)
N
( , ,T)/
( ,T)
N
( , ,T)/

Figure 5.7. Perpendicular Component of 304 Stainless Steel Relative Directional


Spectral Emissivity Coefficients. Rolling experimental (left) and model output (right)
both show little to no change across wavelength.

The Monte Carlo sampling methodology presented previously is followed for


several temperatures for the case of stainless steel 304 and Tungsten. Model output
is compared to the work of Rosenberg (Figure 5.8). Excellent agreement is achieved
between the two methods implying practical convergence in the ideal scenario of a
flat, homogenous surface.
95

0.3
SS304, Model
0.25 SS304, Rosenberg

0.2
W, Model
0.15 W, Rosenberg

0.1

0.05

0
0 1000 2000 3000 4000
Temperature (K)
Figure 5.8. Total hemispherical emissivity coefficients of SS 304 and Tungsten.
Values are generated through Monte Carlo- sampled GOA for a smooth surface and
are compared to the output of Rosenberg.

The model output resembles previous experimental results and converges to


closed-form methods in the case of smooth surfaces. From here, two additional
pieces are pursued: 1) The effect of surface roughness and 2) The discrepancy of
the model at high incident angles.

5.2.3 Model setup for roughened surfaces

The model output resembles previous experimental results and converges to


closed-form methods in the case of smooth surfaces. From here, two additional
pieces are pursued: 1) The effect of surface roughness and 2) The discrepancy of
the model at high incident angles.
96

5.2.3.1 Pseudorandom surface profile generation


The surface profile consists of the sum of 30 cosine waves according to
Equation 5.14 where each sine wave has pseudorandomly generated parameters of
phase shift, amplitude, and (inverse) period.

 ∑   (5.14)

Where

  ,

  ,

  ,

Where all rij are pseudorandomly sampled values from a normal distribution between
0 and 1, and α* is defined by β* and Equation 5.15

500

Traditionally, the surface generation typically used in directional spectral


emissivity modeling techniques involves pseudorandom Gaussian profiles. However,
the emphasis of this piece is to demonstrate the sampling methodology rather than
develop a perfect directional model. For this reason, the Fourier series approach
was taken for the sake of convenience in code development. Adherence to a
Gaussian height distribution is not checked. Each parameter is generated from
pseudorandom sampling of a user-defined range of values having a uniform
probability distribution. Each parameter set (e.g. amplitudes) has the same sampling
range. The roughness of the profile is therefore dictated by the ratio of the amplitude
sampling range to the period sampling range. The significance of this designation is
that the model output is dependent on the relative geometric characteristics of the
surface profile geometry rather than the actual dimensions of the surface profile. It
has been suggested that the slope of the surface profile, rather than the profile
97

amplitude, is the more-significant contributing parameter for emissivity [92] [93].


Furthermore, work in 3-D GOA modeling has emphasized the importance of rms
slope above surface profile height/amplitude [52]. By altering the amplitude –to–
period ratio via the ratio of the parameter sampling ranges, the model falls in
accordance with this trend. To this end, β* is held fixed at an arbitrarily large value to
ensure adequate variation in the sample profile and α* is the independent variable of
the simulation.

Therefore, the value that characterizes the “roughness” of the generated profile is
given as

  (5.15)
For more intuitive profile generation, it may be desirable in the future to
uniformly sample the profile period (1/β*) rather than the inverse period (β*). During
total hemispherical emissivity calculations, a new profile having the same defined Ψ
is generated for each new λ.

5.2.3.2 High-angle scattering


The GOA method was found to consistently overestimate the relative spectral
directional emissivity at high angles in the first two cases. For this reason, a small
correction was included to try and account for the expected inaccuracies of the GOA
method at higher incidence angles.

Originally, the model used the following equation:

, , , , (5.16a)
Where R(λ,θ,T) is calculated through the GOA model. To account for additional
scattering at higher angles, R evaluated by the GOA is replaced.

, , 1 , , (5.16b)

Where
98

, ,   1 , ,

Where

tan 1

And

  .01   .02

This correction only significantly impacts coefficients at high angles of incidence, and
does so while keeping the reflection coefficient between 0 and 1. This correction is
used for the third and fourth cases.

5.2.4 Case 3: Roughened Platinum Surface

It is well established that the GOA approach is limited at high angles of


incidence and for certain roughness-to-wavelength ratios. The two additions of the
model (inclusion of the ζ high angle term and the introduction of a Fourier-series
model profile) give additional quantitative resemblance to spectral directional
emissivity values.

Table 5.2. Normal Spectral Emittance Coefficients


Flat
λ [μm]
Profile (It. 1)
1.5 0.235424 0.235424 0.235432 0.241282537
3 0.100355 0.100355 0.100358 0.103065057
6 0.03455 0.03455 0.034551 0.036974646

At very low Ψ values, the normal spectral emittance values approach the previous
model case of the smooth surface as shown in Table 2. As Ψ increases, normal
spectral emittance values slowly increase, as would be expected from nature.
99

The model is compared to available literature for relative spectral directional


emittance values in both parallel and perpendicular components of light at various
roughness scales for two wavelengths (1.5 and 6 μm). As seen in figures 5.9 and
5.10, the observed trends in the literature are consistent for both wavelengths; the
relative parallel component experiences a decrease from increasing roughness,
observed near peak angles while the perpendicular component experiences an
increase from increasing roughness, observed at approximately the same angle
range. This trend from roughness is resembled in the model output where three
different Ψ values are input, shown in (b) and (d) of figures 5.9 and 5.10.

1.75
(a) (b) = /200
= /40
= /10
( =1.5 m)

1.5

1.25
N
( , =1.5 m)/

0.75
||

0.5
0 15 30 45 60 75 90
(deg)
3.5
=0.2 m
(c) =0.68 m (d)
=1 m
( =6 m)

=2.38 m

2.5
N
( , =6 m)/

1.5
||

0.5
0 15 30 45 60 75 90
(deg)

Figure 5.9. Parallel Component of Emitted Light from (a) Rolling and (b) model
output at wavelengths of 1.5 μm and from (c) Rolling and (d) model output at
wavelengths of 6 μm. Several roughness scales were evaluated.
100

0.5
(a) (b)

( =1.5 m)
0.4
( =1.5 m)

0.3

N
= /200
N

( , =1.5 m)/
= /40
( , =1.5 m)/

= /10
0.2

0.1

0
0 15 30 45 60 75 90
(deg)

(c) (d)
( =6 m)
( =6 m)

N
N

( , =6 m)/
( , =6 m)/

Figure 5.10. Perpendicular Component of Emitted Light from (a) Rolling and (b)
model output at Wavelengths of 1.5 μm and from (c) Rolling and (d) model output
(top) and at Wavelengths of 6 μm. Various roughness scales were evaluated.

Just as in previous cases, the model is Monte Carlo – sampled according to


the Planck Distribution to generate total hemispherical emissivity coefficients across
temperature and the comparison is given in Figure 5.11. According to Equations
5.11a and 5.11b, the high- and low- incidence angle directional emissivity
coefficients have relatively less contribution to the overall emission due to the sin
101

and cos terms. There is relatively little change between the model values in Figures
5.5 and 5.11.
(T)

Figure 5.11. Model and Experimental values [88][89][90][24][91][75] for total


hemispherical emissivity of platinum. Model inputs of Ψ = π/40 and γ=.02 were used.

5.2.5 Case 4: Roughened SS 304 Surface

In the fourth case, the Fourier-series profile is used with the ζ correction factor
while Drude Theory is used to derive the optical constants. This time, there is much
better resemblance between the model and literature in regards to the parallel-
polarized component of relative emitted light from polished stainless steel 304
compared to case 2. The results are provided in Figure 5.12. The perpendicular
component remained unchanged between the two cases.
102
2.5
2.5
=6 m
=6 m
=3 m
=3 m
=1.5 m
=1.5mum
2 2
( ,T)
N

1.5
( , ,T)/

1.5
||

1 1

0.5 0.5
0 15 30 45 60 75 90 0 15 30 45 60 75 90
(deg) (deg)

Figure 5.12. Parallel Component of Relative Spectral Directional Emittance of


polished stainless steel 304 at 953 °K (left) and model output where T = 953 °K, Ψ =
π/200, and γ = .01 (right). The perpendicular component remained unchanged from
case 2.

Adequate spectral directional data could not be obtained for a variety of


roughness scales. Due to the oxidative nature of stainless steel 304 at high
temperatures, only relative total directional emittance data for samples of varying
roughness across multiple temperatures could be found and used as a benchmark,
rather than relative spectral directional. Monte Carlo sampling of wavelengths was
employed to generate total directional emissivity coefficients and compared with
literature values in Figure 5.13.
( ,T)/ (T) ( ,T)/ (T) ( ,T)/ (T)

(c)
N N N (a)

( ,T)/ (T)
(d)
N
(b)
103
104

Figure 5.13. Total Directional Emittance of 304 Stainless Steel. Experimental results
from Rolling are shown in (a) and (c) while model output is provided in (b) and (d).

The Monte Carlo sampling methodology is followed again for several


temperatures for the case of stainless steel 304 and tungsten. Model output is
compared to various sources of experimental literature in figure 14 for stainless steel
304 and figure 15 for tungsten.

0.24
Model, = /10
0.22 Model, = /200
Rolling, =.51 m
0.2 Rolling, =.33 m
Rogers

0.18
(T)

0.16

0.14

0.12

0.1

0.08
200 400 600 800 1000 1200
Temperature (K)

Figure 5.14. SS 304 Total Hemispherical Emissivity Coefficients from


Experiment[75][94]and Model Output
105
0.35
(a) Model, = /10
Model, = /200
Allen
0.3 Forsythe
Matsumoto
Verret
Wojcik
0.25 Worthing

0.2
(T)

0.15

0.1

0.05

0
0 500 1000 1500 2000 2500 3000 3500
Temperature (K)

(b)
(T)

Figure 5.15. (a) Tungsten Total Hemispherical Emissivity Coefficients from


Experimental[95][96][97][98][99][100] and Model Output and (b) Zoomed-in
Tungsten Total Hemispherical Emissivity Coefficients from Experimental[96][98][99]
and Model Output
106

Relatively little difference is apparent between model outputs of case 2


(featured in figure 8) and output of case 4 (figures 14 and 15). The difference in the
two models mostly comes from values at high incidence angles at the directional
spectral level (which are less important due to the sin and cos terms from equations
11a and 11b).

Roughness plays a significant role in overall emissivity. In this study, it was


demonstrated that a spectral directional model where roughness is physically
accounted for can be sampled and iterated into generating a total hemispherical
emittance coefficient where the associated dependencies are still maintained. The
surfaces used in this study were not particularly rough, however, the same process
should be applicable to other techniques where roughness plays a greater role.

5.2.6 Case 5: Oxidized/Roughened SA508 Surface

In this case, to describe the spectral emissivity on oxidized surfaces, we measured


the optical constants for each surface condition, and these constants are
implemented in the code. The Fourier-series profile is also applied with the ζ
correction factor for roughened surfaces. For this case, we picked some SA508
samples of which spectral emissivities were measured with a SOC-100 HDR w/
Nicolet FTIR (Reflective Radiometry). Refractive index (n) and extinction coefficient
(k) were measured by a J. A. Woollham Mark II IR-VASE Ellipsometer. During
measurements, the incident light separates into rays of different path lengths, which
reflect from top and bottom interfaces of the film. The resulting phase differences
produce constructive or destructive interference. Although the ellipsometer does
have limitations for very rough surfaces and thick films, the roughness and film
thickness of all samples evaluated within this study were within the acceptable range
of the measurement method.
107
(a) (b)

Figure 5. 16. Measured spectral optical constants: (a) refractive index and (b)
distinction coefficient.

As shown in Fig. 5. 16, surface oxidation affects both refractive index and extinction
coefficient. It was found that the oxidized surfaces had decreased refractive indices
compared to those of the unoxidized surfaces. This is supported by the conceptual
behaviour of electromagnetic waves propagating through media having bound
electrons versus propagation through media having unbound electrons. It can be
seen that of the real and imaginary components of the refractive index, the real
component is the most sensitive to surface oxidation. Even for the sample of
shortest exposure time (10 hours), there is sharp decrease in refractive index, and
the difference between the 10-hour exposure case and 200-hour exposure case at
350 oC is not significant. In other words, even a thin oxide layer can affect refractive
index, and the refractive indices on oxidized SA508 samples will be similar whether
the oxidation layer is thick or thin in the range examined in this study. There is a
decrease in extinction coefficient on oxidized samples as well. On the contrary, the
extinction coefficient shows gradual change as oxidation time and temperature
increase. While n remained relatively constant after initial oxidation, k continued to
vary for different exposure conditions. We believe that the gradual change in
extinction coefficient would explain the difference of spectral emissivity on samples
oxidized under different conditions. The effect of oxidation on optical constants is
included in our computational model.
108

The comparison shown in Fig. 5. 17 indicates good agreement between


experimental and simulation results. The results indicate that the computational
model based on Kirchhoff’s law can effectively describe spectral emissivity on
sufficiently oxidized surfaces with appropriate optical constant inputs. The model
effectively mimics the experimentally measured spectral emissivity. In metals
(electrically conducting material), the free electrons of the metal can move freely,
which induces large extinction coefficients and phase velocity reduction of wave
propagated in the metal. In the Lorentz model, for metals, electrons and their nuclei
may be treated as simple harmonic oscillators with zero restoring force. However, in
oxide layers (dielectric material), the binding force between electrons and their
respective nuclei, should be considered as a restoring force, which results in a
smaller refractive index and extinction coefficient than the un-oxidized metal surface.
In our computational model, optical constants are represented by the experimentally
measured optical constants for the oxidized surfaces. The bare reference SA508
surface is expected to have an exceedingly thin native oxide layer due to exposure
to air environment even at room temperature.
(a) (b)

Figure 5. 17. Comparison of experimentally measured spectral emissivity and


predicted emissivity by the computational model for (a) mirror polished reference
and oxidized samples at 350 oC and (b) oxidized samples at 300 oC and 600 oC.
109

Therefore, the optical characteristics of the bare reference are mainly attributed to
the pure metal substrate. However, as the oxide layer thickness increases due to
thermal exposure, the proportion of free electrons decreases, consequently reducing
optical constant values. The range of wavelength in which emissivity is affected by
oxidation is expected to be related to the effective penetration length scale of the
wave on the oxidation layer. Even for the heavily oxidized sample (exposed at 600
o
C for 5 hours) where a multi-layered oxide formed, our simulation model could
effectively predict experimentally measured spectral emissivity.
On roughened surfaces, no optical constant change occurs, but the roughened
surfaces lead to more contact of incident light reflected on the surface in the model,
which can explain spectral emissivity increases on the roughened surfaces (Fig. 5.
18). In the model, the surface profile is superimposed by using Fourier series with
selected cosine functions for target roughness which could reproduce trends in
emittance phenomena on the roughened surfaces. In the simulation, we can confirm
that the roughness effect on spectral emissivity is not significant compared to the
effect of oxidation in the range investigated in this study.

Figure 5. 18. Comparison of experimentally measured spectral emissivity and


predicted emissivity by the computational model for roughened surfaces.
110

6 Conclusion 
The goal of studying the emissivity of reactor materials is to develop a flexible,
convenient and accurate theoretical capability of predicting emission coefficients
throughout the reactor system. Experimental measurements of the emission of
several reactor relevant materials in air and relevant Gen IV environments have
been performed. Modeling to predict emissivity has been performed for surfaces of
platinum and tungsten, and AISI 304 stainless steel and the results of these
modeling efforts are in good agreement with existing experimental literature values.
Much of the modeling work is transferable to reactor relevant environments.

6.1 Experimental I: RPV Materials

Spectral directional emittance coefficients of two RPV-candidate materials


were evaluated against multiple dependencies of interest. Oxidation and roughness
were evaluated for SA508 steel (reactor pressure vessel, RPV) while oxidation,
roughness and temperature were evaluated for T91 ferritic steel (a common reactor
structural material intended for RPV for Gen IV reactors).

Two different criteria were successfully evaluated for SA508. In the case of
roughness, significant differences in emittance coefficients were observed only after
samples attained a roughness value greater than that which would be expected of
an operating reactor pressure vessel. In the case of oxidation, SA508 was exposed
to temperatures exceeding steady state LWR, but for a shorter duration than long-
term operation. Samples readily oxidized and emittance coefficient profiles followed
the progression reported in previous literature for pure iron. An attempt at isolating
the effects of temperature was made, however, the surface experienced additional
oxidation during measurement. It is expected that the temperature effects should be
similar to other steels. It was concluded that roughness and temperature should
have little influence on emittance in operating conditions while oxidation at higher
temperature should have a significant effect.
111

Using the data gathered in this study as well as data from previous literature,
approximations for total emissivity coefficients were generated with upper and lower
bounds calculated for each approximation. A total hemispherical emissivity
coefficient above 0.8 for an SA508 reactor pressure vessel may be too generous for
the purpose of conservative heat transfer simulations. Based on this research, the
coefficient of the surface of an in-service RPV approaching long term operation
could be as low as 0.5. For long-term operation, a total hemispherical emittance
coefficient likely lies between 0.6 and 0.85 if it can be assumed that hemispherical
values behave similarly to normal and near-normal values. These values will be
strongly influenced by oxidation and so higher operation temperatures and
durations, which promote oxidation, will mean higher emissivity coefficients.
Temperature-dependent shifts in the Planck Distribution do not affect spectral
emittance but affect total emittance. These will have some significance as long as
there is variation in emittance across wavelengths, as would be the case for the 350
°C SA508 data before 100 hours of exposure (and for T91 in general). However, in
the case of a heavily oxidized surface, effects of Planck shifts are negligible.
Temperature-dependent shifts of the optical constants of the material, or optical
shifts, may indeed occur and could possibly explain the differences between our
data and data in previous literature within the range of 2 to 9 µm. However, this
difference is relatively small (~10%) compared to the overall effect of oxidation.

In the case of T91, it is expected that roughness should also have little
influence on emittance during operating conditions and that oxidation will have far
less of a role than in the case of SA508. This is attributed to the higher Cr content of
T91 which leads to a thin oxide layer at the surface. Optical shifts have been
observed, but these also appear to be relatively minor. For a convenient
conservatively-low approximation of the spectral emittance coefficients of the RPV
surface, it is reasonable to assume coefficients equal to that of an unoxidized, as-
received surface at room temperature since the factors investigated such as
roughness, oxidation and temperature did not have a large significance in the
spectral emittance through typical operation conditions. While in the case of SA508,
112

the change in spectral emittance values increase significantly due to oxidation, this
is clearly not the case for T91.

Planck shifts are reasonably significant for total emittance of a T91 surface.
Neglecting optical shifts, the total emittance coefficient of an unoxidized T91 surface
at 600 °C should be about 20% greater than the total emittance coefficient for the
same surface at 300 °C. Any surface oxidation of T91 that occurs within the realm of
RPV conditions should also increase the significance of Planck shifts. Optical shifts
between 25 °C and 300 °C result in noticeable but minor effects on total emittance of
T91.

6.2 Experimental II: Structural Alloys

The spectral emissivity of structural materials for advanced reactor concepts


has been measured after exposure to relevant environments and temperatures.
Measurements were performed at normal incidence during emission measurement
and at 15-degree incidence during reflectance measurement. The nature and extent
of surface corrosion in these environments has a significant effect on spectral
emissivity.

For Ni – based alloys Inconel 617 and Haynes 230 being considered for high
temperature helium gas cooled reactor (HTGR), a thick black oxide layer develops
on the surface after high-temperature exposure to an oxidative helium gas
environment. This yields a large increase in emissivity. In contrast, the carburizing
He environment yields only a moderate increase in emissivity. Similar effects are
observed in high temperature SC-CO2 environments (of relevance to SC-CO2
Brayton cycle power conversion system common to all advanced reactor concepts),
where oxidation-resistant alloy Haynes 230 exhibits a lower increase in emissivity
than the less oxidation resistant 316 stainless steel. For molten fluoride salt
exposures (coolant for fluoride salt-cooled high temperature reactor, FHR), no oxide
layer forms, but rather corrosion occurs by the dissolution of Cr from the alloy into
the molten salt. Consequently, the lower Cr Hastelloy-N exhibits very little change in
113

emissivity after corrosion tests, while the higher Cr containing 316 stainless steel
exhibits a more notable increase in emissivity. Finally, exposure of 316 stainless
steel in liquid sodium results in negligible change in emissivity because of the low
corrosivity of this environment.

6.3 Theoretical Modeling

The spectral directional GOA model was used to model emissivity of materials
as a function of wavelength, direction, roughness and temperature. While the brunt
of previous modeling research has focused on spectral directional emissivity
(particularly in the fields of optics and metallurgy), total emissivity is obtained from
the spectral directional modeling in this work. Total hemispherical emissivity values
are generated from the simulation of the sampled wavelengths and reasonable
agreement to experimental literature is achieved. The sampling methodology used to
obtain total emissivity presented here through the GOA should be transferrable to
other more-sophisticated emissivity modeling methods such as FDTD. The modeling
progress made in this work demonstrates that first-principles modeling could
potentially be used to model the emissivity of the complex surfaces of a nuclear
reactor.
114

Appendix A – Comprehensive Sample Matrix


Several of the following appendices contain extraneous data sets. The main
experimental section of this piece presents only the data from the “best” method
available to each sample. Generally, the NREL FTIR was considered the best
method followed by Bruker Emittance, followed by Bruker Reflectance, followed by
ET100. Some samples (sodium and salt exposed) were too small to fill the spot size
of the NREL FTIR or the ET100 but could still be cut and “puzzled” for these
measurements. There are several more references in this complete set and as a
result, slightly different ID#’s may be used in the appendices than what is presented
in the main experimental section. The comprehensive project sample matrix is
presented below, ID# designations in the appendices follow this matrix.
Abbreviations follow BE = Bruker Emittance and BR = Bruker Reflectance.

Environ- Sample Sample Profilomtery


Conditions Utilized Systems of Measurement
ment Material ID# Results
Ra RMS ET- Bruker Bruker NREL
[μm] [μm] 100 Reflectance Emittance FTIR
Air-
AE-A-1 Mirror Finish 0.0085 0.009 YES YES NO YES
Exposed
Mirror Finish,
AE-A-2 0.377 0.457 YES YES NO NO
600C, 1 Hr
Mirror Finish,
AE-A-3 0.485 0.621 YES YES NO NO
600C, 3 Hr
Mirror Finish,
AE-A-4 0.602 0.849 YES YES NO YES
600C, 5 Hr
AE-A-5 Shot Peen 2.347 1.882 YES YES NO YES
AE-A-6 120 Grit 0.232 0.293 YES YES NO YES
AE-A-7 320 Grit 0.112 0.145 YES YES NO YES
AE-A-8 800 Grit 0.012 0.018 YES YES NO YES
320 Grit,
SA-508 AE-A-9 0.418 0.594 YES YES NO NO
600C, 3 Hr
AE-A- 800 Grit,
0.336 0.451 YES YES NO NO
10 600C, 3 Hr
AE-A- Mirror Finish,
0.0157 0.03 NO YES NO YES
12 350C, 100 Hr
AE-A- Mirror Finish,
0.016 0.0233 NO YES NO YES
13 350C, 200 Hr
AE-A- Mirror Finish,
0.008 0.0123 NO YES NO YES
14 350C, 10 Hr
AE-A- Mirror Finish,
0.01 0.0137 NO YES NO YES
15 350C, 50 Hr
AE-A- Mirror Finish,
0.0185 0.03 NO YES NO YES
16 300C, 100 Hr
AE-A- Mirror Finish,
0.0163 0.0257 NO YES NO YES
17 350C, 150 Hr

AE-B-1 Mirror Finish 0.007 0.012 YES YES NO YES


Mirror Finish,
T-91 AE-B-2 0.0067 0.0137 YES YES NO NO
750C, 1 Hr
Mirror Finish,
AE-B-3 0.0073 0.0147 YES YES NO NO
750C, 3 Hr
115
Mirror Finish,
AE-B-4 0.0083 0.0513 YES YES NO YES
750C, 5 Hr
AE-B-5 320 Grit 0.059 0.088 YES YES NO YES
320 Grit,
AE-B-6 0.053 0.079 YES YES NO NO
750C, 3 Hr
AE-B-7 Sandblast 1.745 2.21 YES YES NO YES

sCO2 SC-A-1 800 Grit 0.021 0.067 YES NO YES YES


Haynes-230 800 Grit,
SC-A-2 0.26 0.402 YES NO YES YES
650C, 400 Hr

SC-B-1 800 Grit 0.0157 0.0263 YES NO YES YES


SS-316 800 Grit,
SC-B-2 1.234 1.624 YES NO YES YES
650C, 400 Hr

Liquid Na NA-A-1 800 Grit, EDM YES YES YES YES


800 Grit,
SS-316 NA-A-2 0.241 0.3233 YES YES YES YES
650C, 1000 Hr
800 Grit,
NA-A-3 0.2503 0.344 YES YES YES YES
650C, 1000 Hr

Molten
MS-A-1 800 Grit 0.0103 0.0203 YES NO YES YES
Salt
Hastelloy
MS-A-2 800 Grit, EDM 0.0197 0.0623 YES YES YES YES
(FLiNaK)
800 Grit,
MS-A-3 1.273 1.499 YES YES YES YES
850 C, 1000 Hr

MS-B-1 800 Grit, EDM 0.0137 0.0263 YES YES YES YES
SS-316
(FLiNaK) 800 Grit,
MS-B-2 0.3887 0.5093 YES YES YES YES
850C, 1000Hr

750C, 1000 Hr,


MS-C-1 Graphite 0.213 0.303 NO YES NO NO
SS-316
crucible
(FLiBe)
750C, 1000 Hr,
MS-C-2 0.145 0.198 NO YES NO NO
SS316 crucible

He Gas HE-A-1 800 Grit 0.0183 0.0287 NO YES YES NO


1000C, 500 Hr,
Haynes 230 HE-A-2 Carburizing 1.515 1.856 NO YES YES NO
Env.
1000C, 500 Hr,
HE-A-3 0.7297 0.941 NO YES YES NO
Oxidizing Env.

HE-B-1 800 Grit 0.015 0.0223 YES YES YES NO


1000C, 500 Hr,
Inconel 617 HE-B-2 Carburizing 1.424 1.705 NO YES YES NO
Env.
1000C, 500 Hr,
HE-B-3 1.24 1.632 NO YES YES NO
Oxidizing Env.
116

Appendix B: Comparison of Radiometric Reflection Data


sets
This section quickly presents data sets for samples measured using at least the
Bruker Reflectance and the NREL FTIR methods. Samples are sequenced by
roughness (another set of Profilometry measurements was taken for the mirror-
finished SA508 reference). Abbreviations follow BE = Bruker Emittance and BR =
Bruker Reflectance. For additional information on each sample, please refer to the
table in Appendix A. For additional data obtained via the ET-100 system, please
refer to our recent conference proceeding.[87]

Sample: AE-A-1 RMS: 0.0085 RA: 0.009

0.35

0.3

0.25 ET100

0.2 NREL FTIR


ε(λ,θ,T)

BR 1
0.15 BR2
0.1 BR 3
BR 4
0.05
BR avg
0
0 5 10 15 20 25 30
Wavelength [µm]
117

AE-B-1 0.007 0.012

0.4

0.35

0.3
ET-100
0.25
NREL FTIR
ε(λ,θ,T)

0.2 BR 1
BR 2
0.15
BR 3
0.1 BR 4
BR avg
0.05

0
0 5 10 15 20 25 30
Wavelength [µm]

AE-A-14 0.008 0.0123

0.9

0.8

0.7

0.6 NREL FTIR


ε(λ,θ,T)

BR 1
0.5
BR 2
0.4 BR 3
0.3 BR 4

0.2 BR avg

0.1

0
0 5 10 15 20 25 30
Wavelength [µm]
118

AE-A-15 0.01 0.0137

1
0.9
0.8
0.7
NREL FTIR
0.6
ε(λ,θ,T)

BR 1
0.5
BR 2
0.4
BR 3
0.3 BR 4
0.2 BR avg

0.1
0
0 5 10 15 20 25 30
Wavelength [μm]

AE-A-13 0.016 0.0233

0.9

0.8

0.7

0.6 NREL FTIR


ε(λ,θ,T)

BR 1
0.5
BR 2
0.4 BR 3
0.3 BR 4
BR avg
0.2

0.1

0
0 5 10 15 20 25 30
Wavelength [μm]
119

AE-A-17 0.0163 0.0257

0.9

0.8

0.7
NREL FTIR
0.6
ε(λ,θ,T)

BR 1
0.5
BR 2
0.4 BR 3
0.3 BR 4
0.2 BR avg

0.1

0
0 5 10 15 20 25 30
Wavelength [μm]

AE-A-12 0.0157 0.03

0.9

0.8

0.7
NREL FTIR
0.6
ε(λ,θ,T)

BR 1
0.5
BR 2
0.4
BR 3
0.3 BR 4
0.2 BR avg

0.1

0
0 5 10 15 20 25 30
Wavelength [μm]
120

AE-B-4 0.0083 0.0513

1
0.9
0.8
0.7 ET-100
0.6 NREL FTIR
ε(λ,θ,T)

0.5 BR 1
0.4 BR 2
BR 3
0.3
BR 4
0.2
BR avg
0.1
0
0 5 10 15 20 25 30
Wavelength [μm]

AE-B-5 0.059 0.088

0.5

0.45

0.4

0.35
ET-100
0.3 NREL FTIR
ε(λ,θ,T)

0.25 BR 1

0.2 BR 2
BR 3
0.15
BR 4
0.1
BR avg
0.05

0
0 5 10 15 20 25 30
Wavelength [μm]
121

AE-A-6 0.232 0.293

0.6

0.5

0.4 ET100
NREL FTIR
ε(λ,θ,T)

0.3 BR 1
BR 2
0.2 BR 3
BR 4
0.1
BR avg

0
0 5 10 15 20 25 30
Wavelength [μm]

*AE-A-6 is the roughest consensus sample (RCS)

AE-A-4 0.602 0.849

0.9

0.8

0.7
ET100
0.6 NREL FTIR
ε(λ,θ,T)

0.5 BR 1

0.4 BR 2
BR 3
0.3
BR 4
0.2
BR avg
0.1

0
0 5 10 15 20 25 30
Wavelength [μm]
122

AE-A-5 2.347 1.882

0.9

0.8

0.7
ET100
0.6 NREL FTIR
ε(λ,θ,T)

0.5 BR 1

0.4 BR 2
BR 3
0.3
BR 4
0.2
BR avg
0.1

0
0 5 10 15 20 25 30
Wavelength [μm]

AE-B-7 1.745 2.21

0.9

0.8

0.7
ET-100
0.6 NREL FTIR
ε(λ,θ,T)

0.5 BR 1

0.4 BR 2
BR 3
0.3
BR 4
0.2 BR avg
0.1

0
0 5 10 15 20 25 30
Wavelength [μm]
123

Appendix C: Bruker Emittance vs Other Methods


This section quickly presents data sets for samples measured the Bruker
Emittance method and at least one other established method. Samples are
sequenced by roughness (another set of Profilometry measurements was taken for
the mirror-finished SA508 reference), however roughness was only ever a very
significant issue for the Bruker Reflectance Method. Abbreviations follow BE =
Bruker Emittance and BR = Bruker Reflectance. For additional information on each
sample, please refer to the table in Appendix A.

Sample: MS-A-1 RMS: 0.0203 RA: 0.0103

1
0.9
0.8
0.7
0.6 BE T=100C
ε(λ,θ,T)

0.5 BE T=200C
0.4 BE T=300C

0.3 BR avg
NREL FTIR
0.2
0.1
0
1 3 5 7 9 11 13 15 17 19 21 23 25
Wavelength [μm]
124

MS-B-1 0.0263 0.0137

1
0.9
0.8
0.7
0.6 BE T=100C
ε(λ,θ,T)

0.5 BE T=200C
0.4 BE T=300C

0.3 BR
NREL FTIR
0.2
0.1
0
1 3 5 7 9 11 13 15 17 19 21 23 25
Wavelength [μm]

SC-A-1 0.067 0.021

0.9

0.8

0.7

0.6
ε(λ,θ,T)

BE T=100C
0.5
BE T=200C
0.4 BE T=300C
0.3 NREL FTIR

0.2

0.1

0
1 3 5 7 9 11 13 15 17 19 21 23 25
Wavelength [μm]
125

SC-B-1 0.0263 0.0157

0.4

0.35

0.3

0.25
ε(λ,θ,T)

BE T=300C
0.2
BR avg
0.15 NREL FTIR

0.1 ET100

0.05

0
1 3 5 7 9 11 13 15 17 19 21 23 25
Wavelength [μm]

SC-B-2 1.624 1.234

0.9

0.8

0.7

0.6
ε(λ,θ,T)

BE T =100C
0.5
BE T=200C
0.4 BE T=300C
0.3 NREL FTIR

0.2

0.1

0
1 3 5 7 9 11 13 15 17 19 21 23 25
Wavelength [μm]
126

References 

[1] A.J. Gordon, K.L. Walton, T.K. Ghosh, S.K. Loyalka, D.S. Viswanath, R. V.
Tompson, Hemispherical total emissivity of Hastelloy N with different surface
conditions, J. Nucl. Mater. 426 (2012) 85–95.
doi:10.1016/j.jnucmat.2012.03.026.

[2] G. Cao, M.H. Anderson, K. Sridharan, T.R. Allen, In situ measurements of


spectral emissivity of materials for very high temperature reactors nuclear
systems, Nucl. Technol. 175 (2011) 460–467.

[3] R. Maynard, Total Hemispherical Emissivity of Potential Structural Materials


for Very High Temperature Reactor Systems: Hastelloy X, Nucl. Technol. 5450
(2010) 88–100.

[4] R. Maynard, Mokgalap, Hemispherical Total Emissivity of Potential Structural


Materials for Very High Temperature Reactor Systems: Haynes 230, Nucl.
Technol. 179 (2011) 429–438.

[5] S. Weber, Spectral Emissivity Studies for VHTR Candidate Materials,


University of Wisconsin - Madison, 2010.

[6] T.L. Malaney, Spectral Emissivity Measurement and Modeling of High


Temperature Reactor Materials, University of Wisconsin - Madison, 2009.

[7] L. Capone, Y.A. Hassan, R. Vaghetto, Reactor cavity cooling system (Rccs)
experimental characterization, Nucl. Eng. Des. 241 (2011) 4775–4782.
doi:10.1016/j.nucengdes.2011.07.043.

[8] S. Ball, Sensitivity Studies of Modular High-Temperature Gas-Cooled Reactor


( MHTGR ) Postulated Accidents Sensitivity Studies of Modular High-
Temperature Gas-Cooled Reactor ( MHTGR ) Postulated Accidents, (2004) 0–
14.
127

[9] H.D. Gougar, C.B. Davis, Reactor Pressure Vessel Temperature Analysis For
Prismatic And Pebble-Bed VHTR Designs, Idaho Falls, 2006.

[10] C.-D. Wen, I. Mudawar, Modeling the effects of surface roughness on the
emissivity of aluminum alloys, Int. J. Heat Mass Transf. 49 (2006) 4279–4289.
doi:10.1016/j.ijheatmasstransfer.2006.04.037.

[11] T. Iuchi, T. Furukawa, S. Wada, Emissivity modeling of metals during the


growth of oxide film and comparison of the model with experimental results.,
Appl. Opt. 42 (2003) 2317–2326. doi:10.1364/AO.42.002317.

[12] G. Neuer, F. Guntert, In-situ Measurements of Layer Thickness During


Oxidation of Titanium, Thermochim. Acta. (1988) 299–304.

[13] T. Makino, H. Wakabayashi, Thermal Radiation Spectroscopy Diagnosis for


Temperature and Microstructure of Surfaces, JSME Int. J. 46 (2003) 500–509.

[14] L. del Campo, R.B. Pérez-Sáez, M.J. Tello, Iron oxidation kinetics study by
using infrared spectral emissivity measurements below 570 C, Corros. Sci. 50
(2008) 194–199. doi:10.1016/j.corsci.2007.05.029.

[15] G. Cao, S.J. Weber, S.O. Martin, K. Sridharan, M.H. Anderson, T.R. Allen,
Spectral emissivity of candidate alloys for very high temperature reactors in
high temperature air environment, J. Nucl. Mater. 441 (2013) 667–673.
doi:10.1016/j.jnucmat.2013.04.083.

[16] J. Ambrosek, M. Anderson, Current Status of Knowledge of the Fluoride Salt


(FLiNaK) Heat Transfer, Nucl. Technol. 165 (2009) 166–173.

[17] S.-D. Hong, C. Yoon, Analysis of Thermal Radiation Heat Transfer on a Very
High Temperature Gas-Cooled Reactor, 2006.

[18] J.R. Tallackson, Thermal Radiation Transfer of Afterheat in MSBR Heat


Exchangers, 1971.

[19] R.K. Maynard, Total Hemispherical Emissivity of Very High Temperature


128

Reactor (VHTR) Candidate Materials: Hastelloy X, Haynes 230 and Alloy 617,
University of Missouri, 2011.

[20] D.A. Pinnow, T.C. Rich, Development of a calorimetric method for making
precision optical absorption measurements., Appl. Opt. 12 (1973) 984.
doi:10.1364/AO.12.000984.

[21] G. Tanda, M. Misale, Measurement of Total Hemispherical Emittance and


Specific Heat of Aluminum and Inconel 718 by a Calorimetric Technique,
Trans. ASME. 128 (2006) 302–306.

[22] T. Fu, P. Tan, M. Zhong, Experimental research on the influence of surface


conditions on the total hemispherical emissivity of iron-based alloys, Exp.
Therm. Fluid Sci. 40 (2012) 159–167.
doi:10.1016/j.expthermflusci.2012.03.001.

[23] D.F. Simmons, C. Fortgang, Using multispectral imaging to measure


temperature profiles and emissivity of large thermionic dispenser cathodes,
Rev. Sci. Instrum. 76 (2005).

[24] P.R.C. Goard, Application of hemispherical surface pyrometers to the


measurement of the emissivity of platinum, J. Sci. Instrum. 43 (1966) 256–
258.

[25] A. Seifter, M. Grover, D.B. Holtkamp, A.J. Iverson, G.D. Stevens, W.D. Turley,
L.R. Veeser, M.D. Wilke, J.A. Young, Emissivity measurements of shocked tin
using a multi-wavelength integrating sphere, J. Appl. Phys. 110 (2011).
doi:10.1063/1.3656429.

[26] M. Krishna, Spectral Emissivity of Ytterbium Oxide-Based Materials for


Application as Selective Emitters in Thermophotovoltaic Devices, Sol. Energy
Mater. Sol. Cells. 59 (1999) 337–348.

[27] S.M. Baumann, Direct Emissivity Measurements of Painted Metals for


Improved Temeprature Estimation During Laser Testing, Air Force Institute of
129

Technology, 2014.

[28] L.R. Koirala, FTIR-Spectroscopic Measurement of Directional Spectral


Emissivities of Microstructured Surfaces FTIR-Spectroscopic Measurement of
Directional Spectral, Helmut - Schmidt University, 2004.

[29] W. Smetana, R. Reicher, A New Measuring Method to Determine Spectral


Emissivity, Meas. Sci. Technol. (1998) 797–802.

[30] R.G. Hering, T.F. Smith, Surface Radiation Properties from Electromagnetic
Theory, Int. J. Heat Mass Transf. 11 (1968) 1567–1571.

[31] W.J. Parker, G.L. Abbott, Theoretical and Experimental Studies of the Total
Emittance of Metals, in: Symp. Therm. Radiat. Solids, 1964: pp. 11–28.

[32] M. Rosenberg, R.D. Smirnov, A.Y. Pigarov, On thermal radiation from fusion
related metals, Fusion Eng. Des. 84 (2009) 38–42.
doi:10.1016/j.fusengdes.2008.08.046.

[33] E. Hagen, H. Rubens, Metallic Reflection, Ann. Phys. 1 (1904) 352–375.

[34] P. Drude, Zur Elektronentheorie der Metalle, Ann. Phys. 306 (1900) 566–613.

[35] C.J. Powell, Analysis of optical and inelastic-electron- scattering data II.
Application to Al, J. Opt. Soc. Am. 60 (1970) 78–93.

[36] R. Brendel, D. Bormann, An infrared dielectric function model for amorphous


solids, J. Appl. Phys. 71 (1992) 1–6.

[37] S. Roberts, Interpretation of the Optical Properties of Metal Surfaces, Phys.


Rev. 100 (1955) 1667–1671.

[38] N.M. Ravindra, S. Abedrabbo, W. Chen, F.M. Tong, A.K. Nanda, A.C.
Speranza, Temperature-Dependent Emissivity of Silicon-Related Materials
and Structures, IEEE Trans. Semicond. Manuf. 11 (1998) 30–39.

[39] R.W. Gilberd, The anomalous skin effect and the optical properties of metals,
J. Phys. F Met. Phys. 12 (1982) 1845–1860.
130

[40] A.D. Rakic, A.B. Djurisic, J.M. Elazar, M.L. Majewski, Optical properties of
metallic films for vertical-cavity optoelectronic devices., Appl. Opt. 37 (1998)
5271–5283. doi:10.1364/AO.37.005271.

[41] J.R. Howell, M.P. Menguc, R. Siegel, Thermal Radiation Heat Transfer, 6th
Ed., 2016.

[42] S.G. Agababov, Effect of the roughness of the surface of a solid body on its
radiation properties and methods for their experimental determination,
Teplofiz. Vysok. Temp. 6 (1968) 78–87.

[43] S.G. Agababov, Effect of the roughness factor on radiation properties of


solids, Teplofiz. Vysok. Temp. (1970) 770–773.

[44] Agababov, Effect of roughness factor on the radiation properties of a solid


body with random roughness, Teplofiz. Vysok. Temp. 13 (1975) 314–317.

[45] E.I. Thorsos, The validity of the Kirchoff approximation for rough surface
scattering using a Gaussian roughness spectrum, J. Acoust. Soc. Am. 83
(1988) 78–92.

[46] H. Davies, The Reflection of Electromagnetic Waves From a Rough Surface,


Proc. Inst. Elec. Eng. Lond. 101 (1954) 209–214.

[47] A. Oskooi, D. Roundy, MEEP: A flexible free-software package for


electromagnetic simulations by the FDTD method, Comput. Phys. Commun.
181 (2010) 687–702.

[48] Z.Z. Group, RCWA Code Package, (n.d.).

[49] K. Tang, R. Dimenna, Regions of validity of the geometric optics


approximation for angular scattering, Int. J. Heat Mass Transf. 40 (1997) 49–
59.

[50] K. Fu, P. f. Hsu, New regime map of the geometric optics approximation for
scattering from random rough surfaces, J. Quant. Spectrosc. Radiat. Transf.
131

109 (2008) 180–188. doi:10.1016/j.jqsrt.2007.08.019.

[51] K. Tang, The geometric optics approximation for reflection from two-
dimensional random rough surfaces, Int. J. Heat Mass Transf. 41 (1998)
2037–2047. doi:10.1016/S0017-9310(97)00227-5.

[52] D. Bergström, J. Powell, The absorption of light by rough metal surfaces—A


three-dimensional ray-tracing analysis, J. Appl. Phys. 103 (2008) 103515.
doi:10.1063/1.2930808.

[53] J. Qiu, W.J. Zhang, L.H. Liu, P. Hsu, L.J. Liu, Reflective properties of randomly
rough surfaces under large incidence angles., J. Opt. Soc. Am. A. Opt. Image
Sci. Vis. 31 (2014) 1251–8. doi:10.1364/JOSAA.31.001251.

[54] H.J. Lee, B.J. Lee, Z.M.Ã. Zhang, Modeling the radiative properties of
semitransparent wafers with rough surfaces and thin-film coatings, J. Quant.
Spectrosc. Radiat. Transf. 93 (2005) 185–194.
doi:10.1016/j.jqsrt.2004.08.021.

[55] H.J. Lee, Z.M. Zhang, Applicability of Phase Ray-tracing method for light
scattering from rough surfaces, J. Thermophys. Heat Transf. 21 (2007) 330–
336.

[56] J. Qiu, Y.T. Wu, Z. Huang, P. Hsu, L. Liu, H. Zhou, A Hybrid Partial Coherence
and Geometry Optics Model of Radiative Property on Coated Rough Surfaces,
J. Heat Transfer. 135 (2013) 0915031–0915036. doi:10.1115/1.4024466.

[57] B.J. Zhang, C.Y. Zhao, GOA with Considering Interference for Reflection from
Random Rough Surface, J. Thermophys. Heat Transf. 27 (2013) 458–464.

[58] J. Howell, Calculation of radiant heat exchange by the Monte Carlo method, in:
Winter Annu. Meet. ASME, 1965.

[59] Section III, Part UCS, in: ASME Codes Stand., 2010.

[60] G.R. Odette, G.E. Lucas, Embrittlement of nuclear reactor pressure vessels, J.
132

Miner. Met. Mater. Soc. 53 (2001) 18–22. doi:10.1007/s11837-001-0081-0.

[61] C. Cabet, F. Roulliard, Corrosion phenomena induced by gases in Generation


IV nuclear reactors, in: Struct. Mater. Gener. IV Nucl. React. 1st Ed., 2016: pp.
75–104.

[62] V. Ignatiev, A. Surenkov, Corrosion phenomena induced by molten salts in


Generation IV nuclear reactors, in: Struct. Mater. Gener. IV Nucl. React. 1st
Ed., 2016: pp. 153–189.

[63] C. Fazio, F. Balbaud, Corrosion phenomena induced by liquid metals in


Generation IV reactors, in: Struct. Mater. Gener. IV Nucl. React. 1st Ed., 2016:
pp. 23–74.

[64] Y. Ahn, S. Jun Bae, Review of supercritical CO2 power cycle technology and
current status of research and development, Nucl. Eng. Technol. 47 (2015)
647–661.

[65] F.H. Southworth, P.E. MacDonald, D.J. Harrell, The Next Generation Nuclear
Plant (NGNP) Project, Idaho National Laboratory (INL), 2003.

[66] R. Wright, J. Simpson, High Temperature Behavior of Candidate VHTR Heat


Exchanger Alloys, in: Proc. 4th Int. Top. Meet. High Temp. React. Technol.,
ASME, Washington, DC, 2008: pp. 75–79.

[67] T. Allen, J. Busby, M. Meyer, D. Petti, Materials challenges for nuclear


systems, Mater. Today. 13 (2010) 14–23. doi:10.1016/S1369-7021(10)70220-
0.

[68] R.S. Sellers, Impact of Reduction-Oxidation Agents on the High Temperature


Corrosion of Materials in LiF-NaF-KF, University of Wisconsin-Madison, 2012.

[69] G. Zheng, B. Kelleher, G. Cao, M. Anderson, T. Allen, K. Sridharan, Corrosion


of 316 stainless steel in high temperature molten Li2BeF4 (FLiBe) salt, J. Nucl.
Mater. 461 (2015) 143–150. doi:10.1016/j.jnucmat.2015.03.004.
133

[70] R. Leggett, Steady State Irradiation Behavior of Mixed Oxide Fuel Pins
Irradiated in EBR-II, in: Int. Conf. Fast Breed. React. Fuel Perform., Monterey,
1978.

[71] M. Hvasta, Designing and Optimizing a Moving Magnet Pump for Liquid
Sodium Systems, University of Wisconsin - Madison, 2013.

[72] J. Mahaffey, D. Adam, Corrosion of Nickel-Base Alloys in Supercritical


Carbon-Dioxide Environment, in: EPRI Int. Conf. Corros. Power Plants, San
Diego, 2015.

[73] G. Angelino, Carbon Dioxide Condensation Cycles for Power Production, J.


Eng. Power. (1968) 287–295.

[74] J. King, Supplementary Material, (2017).

[75] R.E. Rolling, A.I. Funai, Investigation of the effect of surface conditions on the
radiant properties of metals, Pt II, Wright-Patterson Air Force Base, 1967.

[76] L.N. Aksyutov, Normal spectral emissivity of gold, platinum, and tungsten, J.
Eng. Phys. 27 (1974) 913–917.

[77] S.W. Churchill, H.H.S. Chu, Correlating equations for laminar and turbulent
free convection from a vertical plate, Int. J. Heat Mass Transf. 18 (1975)
1323–1329. doi:10.1016/0017-9310(75)90243-4.

[78] K. Mizuno, J. Ishii, A black body absorber from vertically aligned single-walled
carbon nanotubes, Proc. Natl. Acad. Sci. U. S. A. 106 (2009) 6044–6047.

[79] G. Teodorescu, Radiative emissivity of metals and oxidized metals at high


temperature, Auburn, 2007.

[80] P. Yeh, Optical Waves in Layered Media, John Wiley & Sons, 1988.

[81] M.A. Kats, F. Capasso, Optical absorbers based on strong interference in


ultra-thin films, Laser Photon. Rev. 10 (2016) 735–749.

[82] A.L. Smith, The Coblentz Society Desk Book, The Coblentz Scoeity, Kirkwood,
134

MO, 1982.

[83] R. Wright, Kinetics of Gas Reactions and Environmental Degradation in NGNP


Helium, Idaho Falls, 2006.

[84] C. Cabet, A. Terlain, P. Lett, L. Guétaz, J.M. Gentzbittel, High temperature


corrosion of structural materials under gas-cooled reactor helium, Mater.
Corros. 57 (2006) 147–153. doi:10.1002/maco.200503901.

[85] J.H. Weaver, Optical properties of Rh, Pd, Ir, and Pt, Phys. Rev. B. 11 (1975)
1416–1425.

[86] I.F. Silvera, S. Deemyad, Temperature dependence of the emissivity of


platinum in the IR, Rev. Sci. Instrum. (2008) 79–80. doi:10.1063/1.2966394.

[87] A. Seignac, S. Robin, Optical Properties of Thin Films of Pt in the Far


Ultraviolet, Solid State Commun. 11 (1972) 217–219.

[88] A.H. Sully, Some measurements of the total emissivity of metals and pure
refractory oxides and the variation of emissivity with temperature, Br. J. Appl.
Phys. 3 (1952) 97–101.

[89] C. Davisson, J.R. Weeks, The relation between the total thermal emissive
power of a metal and its electrical resistivity, J. Opt. Soc. Am. Rev. Sci.
Instruments. 8 (1924) 581–605.

[90] P.D. Foote, The Emissivity of Metals and Oxides III. The Total Emissivity of
Platinum and the Relation Between Total Emissivity and Resistivity, Bull. Bur.
Stand. 12 (1915) 607–612.

[91] S.C. Jain, V. Narayan, T.C. Goel, Thermal conductivity of metals at high
temperatures by the Jain and Krishnan method III. Platinum, Brit. J. Appl.
Phys. 2 (1969) 109–113. doi:10.1088/0022-3727/2/1/315.

[92] W. Sabuga, R. Todtenhaupt, Effect of roughness on the emissivity of the


precious metals silver , gold , palladium , platinum , rhodium , and iridium,
135

High Temp. - High Press. 33 (2001) 261–269. doi:10.1068/htwu371.

[93] Y. Yang, R. Buckius, Surface Length Scale Contributions to the Directional


and Hemispherical Emissivity and Reflectivity, J. Thermophys. Heat Transf. 9
(1995) 653–659.

[94] C.R. Roger, S.H. Yen, K.G. Ramanathan, Temperature variation of total
hemispherical emissivity of stainless steel AISI 304, J. Opt. Soc. Am. 69
(1979) 1384–1390. doi:10.1364/JOSA.69.001384.

[95] R.D. Allen, L.F.G. Jr, P.L. Jordan, R.D. Allen, L.F. Glasier, P.L. Jordan,
Molybdenum , Tantalum , and Tungsten above 2300 ° K, J. Appl. Phys. 31
(1960) 1382–1387. doi:10.1063/1.1735847.

[96] W. Forsythe, E. Watson, Resistance and radiation of tungsten as a function of


temperature, Josa. 249 (1934) 114–118.

[97] T. Matsumoto, A. Cezairliyan, D. Basak, Hemispherical Total Emissivity of


Niobium , Molybdenum , and Tungsten at High Temperatures Using a
Combined Transient and Brief Steady-State Technique, Int. J. Thermophys. 20
(1999) 943–952.

[98] D.P. Verret, K.G. Ramanathan, Total hemispherical emissivity of tungsten, J.


Opt. Soc. Am. 68 (1978) 1167–1172.

[99] L.A. Wojcik, A.J. Sievers, Total hemispherical emissivity of W (100), J. Opt.
Soc. Am. 70 (1980) 443–450.

[100] W.E. Forsythe, A.G. Worthing, The properties of tungsten and the
characteristics of tungsten lamps, Astrophys. J. 61 (1925) 146–185.
doi:10.1017/CBO9781107415324.004.

[101] J.L. King, H. Jo, G. Cao, Emissivity of Reactor Structural Alloys, 8th
International Topical Meeting on High Temperature Reactor Technology (HTR
2016)

Das könnte Ihnen auch gefallen