Optimization of Transmission Line Measurement (TLM) Structures for Specific

Contact Resistivity Determination

Sidhant Grover, Rochester Institute of Technology1,1 , Department of Electrical & Microelectronic Engineering1

Abstract
Low resistance ohmic contacts are of extreme importance to modern semiconductor devices. As device sizes continue to
shrink, it implies a reduction in the specific contact resistance ρc . There are various methods for the measurement of
ρc , however the Transmission Line Model (TLM) is most popularly used to determine the specific contact resistivity for
Integrated Circuits (IC) and Silicon Photovoltaics (PV) applications. Inconsistencies have been observed in literature
between IC and PV devices as ρc determination may depend on dimensions. Therefore, TLM test geometries need to
optimized in order to minimize error. Optimimum values of TLM widths were fabricated and tested and systematic
error was compared with that from simulations.
Keywords: Sheet Resistance, Contact Resistance, Transmission Line Model (TLM), Specific Contact Resistivity,
Ohmic Contacts, Error Analysis, Optimization

1. Introduction ρc . Simulations carried out in this paper as well as in

Ueng et al [2] suggest optimum values of the widths of
In the modern semiconductor industry, device perfor-
these TLM structures that provide the least uncertainty
mance can by impacted by various means. One of the
in measurement. A contour map similar to one below can
major factors that effect the performance of devices is the
be formulated for optimum TLM contact geometries for
resistance between the contacts and the device itself. This
various values of RSH and ρc yielding to different semi-
is termed as contact resistance. Accurate determination
conductor applications.
of this contact resistance is essential in understanding its
impact on device performance. Tremendous reduction in ρc
device sizes in modern times have require specific contact
resistivity (ρc ) improvement to maintain small parasitic re- ρc2 Wopt2
sistances with small contact area within acceptable ranges.
These parasitic resistances need to be significantly small ρc1 Wopt1
for practically all semiconductor applications. Optimum
pattern designs for the Transmission Line Model (TLM) RSH1 RSH2 RSH
that are suggested to achieve minimum measurement un-
certainty of the specific contact resistance can be devel- Figure 1: Map of optimum widths for different applications 1 and 2.
oped for semiconductor applications of varying ρc and RSH .
The determination of the specific contact resistance The above figure shows a map of the different optimum
and sheet resistance (RSH ) of a planar ohmic contact struc- widths of TLM structures for applications 1 and 2 with
ture and the underlying doped layer can be done through values of sheet resistances RSH1 and RSH2 , and specific
the standard Transmission Line Model (TLM) method.[1] contact resistivities ρc1 and ρc2 .
The contact current injection in a standard TLM structure Conventional TLM structures do not accurately de-
is in a lateral channel that makes it result in resistance val- termine the transfer length LT of the contact. Lateral
ues commensurate to FET as well as photovoltaic device contacts that have contact lengths much larger than LT
structures. These device structures have varying ρc and behave as semi-infinite contacts. As the length of the con-
RSH values and the contacts differ between being alloyed tacts is decreased, below LT , the resistance of the contact
and non-alloyed metal contacts. increases sharply. [3]
There is no information on the uncertainties of these
alloyed and non-alloyed structures and hence one cannot 2. Theory
accurately compare data sets. Similar issues arise for dif-
ferent applications employing varying values of RSH and 2.1. Ohmic Contacts and Contact Resistance
All semiconductor devices have contacts and all con-
URL: www.rit.edu (Rochester Institute of Technology) tacts have contact resistance. Metal-to-Semiconductor con-
Preprint submitted to Elsevier May 19, 2016
34th Annual Microelectronic Engineering Conference, May 2016 2.2 TLM Analysis

tacts have been of increasing importance to the semiconduc- Rc Rc

tor industry in order to better performance of devices by A B
reducing this resistance. A large mismatch in the fermi en- Rsemi n
ergy between the metal and the semiconductor can result
in a high resistance rectifying contact. p
Ohmic contacts are preferred for semiconductor appli-
cations as they have linear or quasi-linear and symmetric
I-V characteristics. The voltage drop at the contact for Figure 3: Two contacts to a diffused semiconductor layer indicating
ohmic contacts is small compared to that of the active the contact resistances and semiconductor resistance.
regions of the devices. Ohmic contacts do not degrade de-
vice performance and do not inject minority carriers. An
ohmic contact results if the Shottky barrier height (φB ) in resistance of the semiconductor is given by Rsemi . The
a metal-semiconductor is zero or negative. Carriers then equation governing the total resistance of the system is
flow freely in and out of the junction so that there is min- given in equation (3).
imal resistance across the contact. The contact resistance
is given as the resistance of the metal-semiconductor junc- 2.2. TLM Analysis
tion and the resistance of the semiconductor material. The The Transmission Line Measurement (TLM) model is
specific contact resistivity is a figure-of-merit for ohmic effective in obtaining the specific contact resistance be-
contacts. It is defined as the slope of the I-V curve when tween a metal and a semiconductor. The current flow in
the voltage equals zero. the semiconductor us uniform but not through the con-
 ∂J  tact. On flowing from the semiconductor to the metal, it
ρc ≡ (1) encounters specific contact resistivity (ρc ) and sheet re-
∂V V =0 sistance (RSH ). Therefore, it chooses the path of least
The specific contact resistivity is a function of the bar- resistance leading to ”current crowding” causing a drop in
rier height as well as the doping density (ND ) as illustrated current from the edge of the contact. The transfer length
below. (LT ) is the average the electron (or hole) travels in the
semiconductor beneath the contact before it flows up into
the contact. The expression for the transfer length is given
by Reeves[1].
r
ρc 
LT = (2)
RSH
The equation for the contact resistance (Rc ) is given
below in equation (2) as the specific contact resistivity can
be written as a product of the sheet resistance and transfer
length
ρc RSH LT
ρc = RSH LT ⇒ Rc = = (3)
LT W W
The total resistance RT is the sum of the resistance of
the semiconductor below the contact and the resistance of
the contact itself.
L RSH LT
RT = Rsemi + 2Rc = RSH +2 (4)
Figure 2: ρC as function of barrier height and ND [4] W W
The above equation is then used to obtain the plot for
From figure 2, it can be observed that for ND ≥ 1019 the extraction of the various resistive values. A standard
cm− 3, ρc is dominated by the tunneling process and de- TLM test pattern includes a single rectangular doped re-
creases rapidly with increased doping. ForND ≤ 1017 gion that signifies a specific sheet resistance (RSH ) value.
cm− 3, the current is due to thermionic emission and ρC is An array of contacts are fabricated with the same width
predominantly independent. In between the two regions, and different pad spacings on that particular doped region.
a combination of tunneling and thermionic emission takes Resistance measurements are then taken on each pair of
place. these contacts that is used to construct the TLM graph
The resistive components that add up in the formation from which RSH , RC , LT and ρC can be determined.
of a contact are illustrated in figure 3. The contact re-
sistance of the two contacts are given as Rc each and the

2
34th Annual Microelectronic Engineering Conference, May 2016 2.4 Error Analysis

RT
2.4. Error Analysis
There are two categories that are used to classify the
error propagation. These are: Random Error and System-
Slope = RSH /Z
atic Error. The difference between a single measurement
of a parameter and the mean determined by a large num-
ber of trials is the Random Error. The consistent shift of
2LT 2RC
means due to taking larger numbers of data points is the
0 d1 d2 d3 d4 d
Systematic Error. The uncertainty in measurement of the
specific contact resistivity ρc is taken into account in this
(a) Plot used to extract Contact Resistance Rc analysis.The equation for the relative uncertainty due to
systematic error is given by the equation below.
Bare Si
MESA

s
4
    
Z W δρc W RSH
d1 d2 d3
= √ δR + δd + δW (7)
ρc ρc RSH ρC W

Here δR, δd, and δW are the measurement uncertain-

ties in the resistance, pad spacing and widths of the TLM
Metal Contacts
structures. The values of W that yield minimum uncer-
(b) Top-down view of TLM structures tainty in ρc is found by evaluating the partial derivative
Insulating Oxide
of equation (4) and equating it to zero. This gives us the
P+ n-type Implant
MESA equation for Wopt .
Bare Si p-type s  
δW
Wopt = 4 ρC RSH (8)
p
(c) Horizontal view of fabricated TLM structure δR
Figure 4: Plot and structure to extract ρc Different combinations of ρc and RSH pertaining to dif-
ferent applications yield particular values of the optimum
width Wopt that gives us the least uncertainty in measure-
2.3. Specific Contact Resistivity (ρc ) extraction
ment. A contour plot for these values of optimum widths
The general equation to extract the specific contact can be generated with respect to ranges of ρc and RSH .
resistivity ρc is given by
!2 3. Experimental Procedure
(RC W ) 1
ρc = ∗ (5)
coth LLT RSH The observed inconsistencies in literature were initially
r attributed to the error in measurements. Upon optimiza-
ρc tion of the equation for the relative uncertainty due to sys-
LT = ( ) (6)
RSH tematic error, the equation for optimum width (Wopt ) did
There are two limiting cases due to the use of the cotan- not depend upon the TLM length (L) or transfer length
gent hyperbolic function. First, is the short contact limit (LT ). Therefore, only TLM widths were varied in the
where the length of the contact is less than half of the experiment. The process was designed based on specific
transfer length. In that case the length of the contact is values of sheet resistance that pertained to PV and IC ap-
used in the ρc calculation. Second, is the long contact limit plications. Contacts with Aluminum and Nickel Silicide
where the length of the contact is significantly greater than (NiSi) were fabricated.
the transfer length. Then the transfer length is used in the A 3 level mask was designed for Mesa lithography,contact
extraction. cut and metal etch processes. TLM length was fixed at
Two limiting cases 10µm and the widths were varied from 10µm and 2000µm.
  There were 4 spacings between he contacts d1 ,d2 , d3 and
• Lc < 0.5LT → coth LLT ≈ LLT ⇒ ρc = Rc W L d4 and they were set at 30µm, 60µm, 120µm and 240µm
respectively.
Short Contact Limit
The fabrication process consists of 6” bare Si p-type
wafers with initial bulk resistivities ranging between 1-
 
• Lc > 1.5LT → coth LLT ≈ 1 ⇒ ρc = Rc W LT
30 Ω/cm3 . These wafers are implanted with n+ dopant
Long Contact Limit (Phosphorous, P31) through a 30nm screening oxide and
annealed in a nitrogen ambient to activate the dopants.
Application specific values of the sheet resistance based

3
34th Annual Microelectronic Engineering Conference, May 2016 4.1 Optimization Results

Figure 6: Contour map from systematic error optimization

value gave an optimum width of 900µm and low RSH gave

an optimum width of 150µm.
Figure 5: Mask design for 100µm width TLM The TLM’s were fabricated and tested to obtain the
specific contact resistivity for each width and the relative
systematic error for each TLM measurement was obtained
on silicon photovoltatics and the CMOS IC applications from equation (7). The values of δd, δR and δW were
were chosen. set at 0.1µm, 0.2Ω and 1.5µm respectively. TLM’s on
• Emitter RSH in silicon photovoltaics - 50 -200 Ω/. Aluminum metal contacts to high sheet resistance were not
analyzed due as schottky contact behavior was observed.
• CMOS IC RSH -1000 - 3000 Ω/. The relative systematic error was plotted as a function of
the optimum width to observe the values of TLM widths
that gave the least amount of error. The values obtained
Wafer Dose(cm−3 ) RSH (Ω/) Metal for each of the wafers were a relative close match to those
1 9.5E12 1500 Aluminum obtained via simulation.
2 9.5E12 1500 NiSi
3 2E15 50 Aluminum
4 2E15 50 NiSi

Table 1: Process design parameters

A dry silicon reactive ion etch process is used to de-

fine MESA patterns on the wafers about 2µm deep. The
MESA etch is spefically deep in order to go below the junc-
tion depth of the implanted dopant to ensure sufficient iso-
lation. TEOS is used as an insulating oxide for the TLM
contact pads. Contact cut openings were patterned using
i-line lithography and the TEOS was etched using a reac-
tive ion etch process. Metal was then deposited and pat-
terned. For the NiSi wafers, the metal was first annealed
in a Rapid Thermal Process (RTP) to form the silicide
and the unreacted Ni was removed using a piranha etch.
Aluminum was then deposited on top of NiSi, patterned Figure 7: Relative systematic error as a function of TLM widths
and etched.
Specific contact resistivity (ρc ) extracted from the TLM
4. Results test measurement of the optimum widths were relatively
similar to anticipated values.
4.1. Optimization Results
Systematic error optimization of equation (7) gave op-
timum width values for application specific ρc and RSH . On measurement of the non-optimum values of the
The contour map as shown in figure (6) is obtained TLM widths, a trend between the TLM width and the ex-
and the values of optimum widths for a range of ρc and tracted transfer length (LT ) was observed. Due to this de-
RSH values. The optimum widths obtained for the RSH pendence, measurement of optimum values of TLM widths
applications used in this paper are shown. The high RSH can give inconsistent results due to incorrect use of the
TLM formula limits.
4
34th Annual Microelectronic Engineering Conference, May 2016

ρc (Ω − cm2 ) RSH formation about the extracted transfer length, which, as

ρc (Ω − cm2 )
Metal Antici- (Ω/sq.) we have seen earlier, depends on the width of the TLM
Extracted
pated Fabricated structure. Figure(9) shows that there is a large delta ob-
NiSi ∼ 10−5 53 2 ∗ 10−5 served in the value of ρc extracted from the use of the gen-
NiSi ∼ 10−3 1530 1 ∗ 10−3 eral formula and the long contact limit.Therefore, while
Al ∼ 10−5 51 4.5 ∗ 10−5 using an optimized value of the TLM width in order to
obtain the least amount of systematic error in measure-
Table 2: Comparison of anticipated and extracted values
ment, the transfer length extracted from that width must
be carefully compared to the contact length and appro-
priate formula must be used. An approach is required to
accurately determine ρc for a given application space.

5. Conclusion

A process is suggested to accurately determine the spe-

cific contact resistivity (ρc ) from the TLM method for a
given application space.

Anticipate ρc and RSH

Chosen value of L

Obtain Wopt from Error Analysis

Figure 8: Transfer length as a function of TLM width

It can be observed from figure (8) that for a transfer Determine LT at Wopt
length extracted from a TLM measurement smaller than
the length of the contact, the long contact approximation
is to be used ans the short contact approximation for trans- LT > 2L LT < 1.5L
fer lengths significantly greater than the length of the con-
tact in order to accurately determine the specific contact Short Contact Limit Long Contact Limit
resistivity.
Figure 10: Suggested process flow to accurately determine ρc

Appropriate application of the long or short contact

approximations for ρc determination by the TLM method
are suggested in figure 10. It is strongly suggested that the
general formula be applied to avoid any inconsistencies in
calculation.

6. Future Work

Further understanding behind the interaction of trans-

fer length and TLM width needs to obtained and the effect
of varying TLM lengths on transfer length and ρc extrac-
tion needs to be investigated. A standardized approach
for accurate determination of the specific contact resis-
tivity (ρc ) and simultaneous comparison with universal
Cross Bridge Kevin Resistance (CBKR) curves needs to
Figure 9: ρc overestimation in long contact limit be developed. Different contact metals and/or metalliza-
tion schemes can also been investigated.
The TLM design can have a significant impact on ρc
determination. The application of the long contact limit
has been repeatedly seen in literature without accurate in-
5
34th Annual Microelectronic Engineering Conference, May 2016

Acknowledgment

The author would like to thank Dr. Santosh Kurinec

for her ever-helpful guidance. The RIT SMFL staff for
equipment maintenance, Sean O’Brien, Patricia Meller and
Microelectronic Engineering MS students Astha Tapriya,Karine
Florent and Alex Marshall for process guidance.

References
[1] G. K. Reeves, H. B. Harrison, Obtaining the specific
contact resistance from transmission line model measure-
ments, IEEE Electron Device Letters 3 (5) (1982) 111–113.
doi:10.1109/EDL.1982.25502.
[2] H.-J. Ueng, D. B. Janes, K. J. Webb, Error analysis leading
to design criteria for transmission line model characterization of
ohmic contacts, IEEE Transactions on Electron Devices 48 (4)
(2001) 758–766. doi:10.1109/16.915721.
[3] D. Sawdai, D. Pavlidis, D. Cui, Enhanced transmission
line model structures for accurate resistance evaluation of
small-size contacts and for more reliable fabrication, IEEE
Transactions on Electron Devices 46 (7) (1999) 1302–1311.
doi:10.1109/16.772468.
[4] T. Schreyer, S. Swirhun, W. Loh, K. Saraswat, R. Swanson, Com-
parison of test structures used for the measurement of low re-
sistive metal-semiconductor contacts, IEEE VLSI Workshop on
Test Structures (1986) 7–23.

Author

Sidhant Grover is an undergraduate

student in the Microelectronic En-
gineering and Materials Science and
Engineering B.S/M.S. program and
at RIT. He has worked intern posi-
tions with the Yield Engineering and
Diffusion Furnace groups at Glob-
alfoundries and Test Engineering at
Analog Devices, Inc., contributing
to cross-disciplinary projects that ef-
fect those companies on a global
scale. Sidhant’s research interests in-
clude contact resistance reduction, Ge
CMOS,III-V,SOI and emerging device
technologies.