R ap i d T h e rm al A n n e al i n g

Introduction
In the semiconductor industry, rapid thermal annealing (RTA) is a semi-conductor
process step used for the activation of dopants and the interfacial reaction of metal
contacts. In principle, the operation involves rapid heating of a wafer from ambient to
approximately 10001500 K. As soon as the wafer reaches this temperature, it is held
there for a few seconds and then finally quenched. A rapid process step is crucial in
order to avoid too much diffusion of the dopants. Furthermore, it is also important to
avoid overheating and nonuniform temperature distribution to occur. An RTA
apparatus uses high-power IR lamps as heat sources (Ref. 1).

oven enclosure
lamp sensor

lamp

wafer

quartz tube

wafer sensor
infrared filter
photo-diode detector

Figure 2-51: Diagram of a typical RTA (rapid thermal annealing) apparatus.

A technical difficulty lies in how to properly measure the wafers temperature during
the process. Two commonly used technical solutions are: thermocouples and IR
sensors.
To achieve an accurate measurement, it is important that the temperature sensor not
be subjected to direct radiation from the lamp. Ideally positioned, the sensor only
receives secondary radiation; that is, the radiation reflected and emitted by the silicon
wafer. Desirable characteristics of the sensor are high accuracy and short response time.
While a high-performance design requires superior electronics, the sensor geometry
plays a big role. In a nutshell, the sensor needs to be large enough to register a
sufficient amount of radiation but light enough to minimize its own thermal inertia.
Since COMSOL Multiphysics gives you control over the geometry, a parameter

136 |

C H A P T E R 2 : E L E C T R O N I C S A N D P O W E R- S Y S T E M M O D E L S

optimization of the sensor could be an exciting project. But first, justify that an infrared
sensor is indeed more appropriate than the inexpensive thermocouple.

Model Definition
Figure 2-51 illustrates a typical RTA configuration. In many applications, RTA makes
use of double-sided heating, in which IR lamps are positioned both above and below
the silicon wafer. In this example we are modeling a single-sided heating apparatus, as
depicted in Figure 2-52.

silicon wafer

sensor
heater

Figure 2-52: The model geometry.

The components in Figure 2-52 are contained in a chamber with
temperature-controlled walls with a set point of 400 K. This results in a closed cavity
so you can omit the geometry of the chamber walls. Furthermore, the model assumes
that this physical system is dominated by radiation and convection cooling. The
convective cooling of the wafer and sensor to the gas (at 400 K) is modeled using a
heat transfer coefficient, h (in this example set to 20 W/(m2K)).
The problem is governed by the heat equation, given below together with its boundary
conditions:
T
C p ------- + ( k T ) = Q
t
4

n ( kT ) = h ( T inf T ) + ( ( 1 ) ) ( J 0 T )

RAPID THERMAL ANNEALING

137

Here is the density; k denotes the thermal conductivity; Q represents the volume heat
source; n is the surface normal vector; Tinf equals the temperature of the convection
cooling gas; denotes the surface emissivity; J0 is the expression for surface radiosity
(further described in the Heat Transfer Module Users Guide); and is the
Stefan-Boltzmann constant.
The model simulates the lamp as a solid object with a volume heat source of 25 kW. It
is insulated on all surfaces except the for the top, which faces the silicon wafer. At this
surface, heat leaves the lamp as radiation only. In order to capture the lamps transient
startup time, the model uses a low heat capacity, Cp, for the solid (10 J/(kgK)). The
lamps other thermal properties are identical to those of copper metal (the default
value in the application mode).
In this case assume that the wafer dissipates energy via radiation and convection on all
surfaces. The sensor is insulated at all surfaces except the top, which is subjected to
both convection and radiation. The thermal material properties are set to those of
alumina.
The following table summarizes the material properties used in the model:
TABLE 2-6: MATERIAL PROPERTIES
MATERIAL

k (W/(mK))

(kg/m3)

Cp (J/(kgK))

IR lamp

400

8700

10

0.99

Silicon wafer (silicon)

163

2330

703

0.5

Sensor

27

2000

500

0.8

The model simulates the transient temperature field for 10 s of heating. The initial
temperature is 400 K for all objects.

138 |

C H A P T E R 2 : E L E C T R O N I C S A N D P O W E R- S Y S T E M M O D E L S

Results and Discussion

Figure 2-53 displays the temperature distribution after 10 s of heating.

Figure 2-53: Temperatures of the lamp, wafer, and sensor after 10 s of heating.
After 10 seconds, the temperatures of the wafer and sensor differ significantly: the
wafer is at 1800 K, whereas the sensor is at 1100 K.
Notice that the temperature distribution in the wafer with a delta of several hundred
degrees is not very uniform, and that you probably can do much better by
reconfiguring the heat source. However, such a reconfiguration is not included in this
model.
To investigate how well the sensors temperature reflects that of the wafer surface, it is
useful to plot the temperature transient of the wafer surfaces centerpoint that faces the

RAPID THERMAL ANNEALING

139

lamp (Twafer), together with the temperature at a point on the sensor top surface
(Tsensor) (see Figure 2-54).

Figure 2-54: The temperature transients of the lamp, the silicon wafer, and the sensor,
together with the irradiation power at the sensor surface.
The sensor temperature reflects that of the silicon wafer poorly. This means that the
signal of a thermocouple, positioned anywhere in the sensor domain of Figure 2-52, is
of little use for regulating this process.
The IR-detector transient (Sensirrad) matches the wafer temperature characteristic
quite well. A scalar amplification allows for a high accuracy measurement of the wafer
temperature. The precise amplification factor is system-dependent and subject to a
calibration requirement.
However, IR-sensor methodology also has drawbacks. The IR signal depends on the
emissivity of the wafer, which will vary with temperature making the response
nonlinear. Furthermore, the IR signal is very sensitive to geometry changes.
The bright side is that COMSOL Multiphysics does not set any limits with respect to
these phenomena and allows you to study them fully.

140 |

C H A P T E R 2 : E L E C T R O N I C S A N D P O W E R- S Y S T E M M O D E L S

Reference
1. A.T. Fiory, Methods in Rapid Thermal Annealing, Proc. 8th Intl Conf. Advanced
Thermal Processing of Semiconductors (RTP 2000), http://web.njit.edu/~fiory/
Papers/RapidThermalAnnealing00.pdf, pp. 1525.

Model Library path: Heat_Transfer_Module/

Electronics_and_Power_Systems/thermal_anneal

Modeling Using the Graphical User Interface

MODEL NAVIGATOR

1 Open the Model Navigator.

2 From the Space dimension list, select 3D.
3 In the list of application modes select Heat Transfer Module>General Heat
Transfer>Transient analysis, then click OK.
CONSTANTS AND EXPRESSIONS

1 From the Options menu, select Constants.

2 Define the following constants; when done, click OK.
NAME

EXPRESSION

DESCRIPTION

T_wall

400[K]

Temperature, wall

T_gas

400[K]

Temperature, gas

h_gas

20[W/(m^2*K)]

Heat transfer coefficient

k_sens

27[W/(m*K)]

Thermal conductivity, sensor

rho_sens

2000[kg/m^3]

Density, sensor

Cp_sens

500[J/(kg*K)]

Heat capacity, sensor

e_sens

0.8

Surface emissivity, sensor

e_lamp

0.99

Surface emissivity, lamp

q_lamp

25[kW]/(pi*50^2*1[mm^3])

Heating power, lamp

e_wafer

0.5

Surface emissivity, wafer

Cp_lamp

10[J/(kg*K)]

Heat capacity, lamp

ampl

50

Amplification factor

RAPID THERMAL ANNEALING

141

GEOMETRY MODELING

1 Create three cylinders. To do so, open the menu item Draw>Cylinder and enter these

settings; when finished, click OK.

OBJECT

RADIUS

HEIGHT

AXIS BASE POINT, X

AXIS BASE POINT, Z

CYL1

0.05

5e-4

CYL2

0.05

1e-3

-5e-2

CYL3

1e-2

1e-3

0.07

-5e-2

2 Click the Zoom Extents button on the Main toolbar.

PHYSICS SETTINGS

Subdomain Settings
1 From the Physics menu, select Subdomain Settings.
2 In the General page, select Subdomain 1. In the Cp edit field type Cp_lamp, and in

the Q edit field type q_lamp. Use the default values for both the conductivity and
the density.
3 Select Subdomain 2, then click the Load button. In the Material list, select Basic
Material Properties>Silicon. Click OK.
4 Select Subdomain 3. In the k edit field type k_sens, in the edit field type
rho_sens, and in the Cp edit field type Cp_sens.

5 Select all subdomains. Click the Init tab, then in the T(t0) edit field type T_wall.
6 Click OK.

Boundary Conditions
1 From the Physics menu, open the Boundary Settings dialog box.
2 Select Boundary 4. In the Boundary condition list select Heat flux, and in the Radiation
type list select Surface-to-surface. In the edit field type e_lamp, and in the Tamb edit

field type T_wall.

3 Select Boundaries 58, 10, 12, and 16.
4 In the Boundary condition list select Heat flux. In the h edit field type h_gas, and in

the Tinf edit field type T_gas.

5 In the Radiation type list select Surface-to-ambient. In the edit field type e_wafer,

and in the Tamb edit field type T_wall.

6 Select Boundaries 7 and 16. From the Radiation type list, select Surface-to-surface.
7 Select Boundary 16 alone. Change the entry in the edit field to e_sens. Click OK.

142 |

C H A P T E R 2 : E L E C T R O N I C S A N D P O W E R- S Y S T E M M O D E L S

MESH GENERATION

1 From the Mesh menu, open the Free Mesh Parameters dialog box.
2 In the Predefined mesh sizes list, select Coarser. Click the Advanced tab. In the
z-direction scale factor edit field type 5. Click the Remesh button, then click OK.
PREPARE POSTPROCESSING

To prepare some postprocessing operations, you need to define an integration

coupling variable. In addition, a line intersecting the lamp and wafer surfaces is also
helpful in later postprocessing.
1 Choose Options>Integration Coupling Variables>Boundary Variables.
2 Select Boundary 16. In the Name edit field type Sens_irrad, and in the Expression

edit field type G_htgh (a predefined application mode variable representing inward
radiation which includes both surface-to-surface and surface-to-ambient
contributions). Click OK.
3 From the Draw menu select Line. In the edit fields for x, y, and z type 0 0, 0 0, and
-5e-2 1e-3, respectively. Click OK. (This step is not necessary if you loaded the

geometry file).
COMPUTING THE SOLUTION

1 From the Solve menu, select Solver Parameters.

2 On the General page, type 0 10 in the Times edit field.
3 In the Linear system solver list, select Direct (UMFPACK).
4 Click the Time Stepping tab. In the Times to store in output list, select Time steps from
solver.
5 From the Consistent initialization of DAE systems list, select On.
6 From the Error estimation strategy list, select Exclude algebraic. Click OK.

The last setting instructs the solver to omit the radiation calculations, which is
always a stationary solution (algebraic equation), from the time-stepping error
analysis. This greatly speeds up the solution process in terms of time stepping.
7 Click the Solve button on the Main toolbar (the solving process should take less than

a minute).
PO STPRO CES SING A ND VISUAL IZATION

Generate Figure 2-53 by executing the following instructions:

1 From the Postprocessing menu open the Plot Parameters dialog box.

RAPID THERMAL ANNEALING

143

2 On the General page, clear the Slice check box and select the Boundary check box in

the Plot type area. Click OK.

The following steps generate Figure 2-54:
1 From the Postprocessing menu, open the Domain Plot Parameters dialog box.
2 On the General page, select the Keep current plot check box.
3 On the Point page, select Point 10.
4 In the y-axis data area, verify that the selection in the Predefined quantities list is
Temperature.
5 Click the Line Settings button. In the Line marker list, select Triangle.
6 Select the Legend check box, then click OK.
7 Click Apply to plot the lamp temperature.
8 Select Point 12, then click the Line Settings button.
9 In the Line marker list select Square, then click OK.
10 Click Apply to add the plot of the wafer temperature.
11 Select Point 23, then click the Line Settings button.
12 In the Line marker list select Circle, then click OK.
13 Click Apply to generate the sensor temperature plot.

Finally, add the IR-detector transient to the plot:

14 In the Expression edit field, type Sens_irrad*ampl.
15 Click the Line Settings button.
16 Set the Line color to Color, the Line style to Dashed line, and the Line marker to None.

Click OK.
17 Click OK to generate the plot and close the Domain Plot Parameters dialog box.
18 Click the Edit Plot toolbar button in the figure window.
19 Select the first Line object in the Axes tree on the left, then type T_lamp

in the Legends dialog box.

20 Repeat the previous step for the remaining three line objects, entering the legends
T_wafer, T_sensor, and Sens_irrad,

respectively.
21 Click OK to close the Edit Plot dialog box and finish the plot.

144 |

C H A P T E R 2 : E L E C T R O N I C S A N D P O W E R- S Y S T E M M O D E L S