You are on page 1of 4

Q-12

MICROWAVE RESISTANCE OF GALLIUM ARSENIDE AND SILICON P-I-N DIODES

ROBERT H. CAVERLY AND GERALD HILLER

SOUTHEASTERN MASS. UNIV. M/A-COM SEMICONDUCTOR PRODUCTS


N. DARTMOUTH, MASS. 02747 BURLINGTON, MASS. 01803

Abstract whert,? W is the j–region width, p is the ambipoiar


mobl lit v, and Q is the stored charge, which is
The purpose of this paper is to demonstrate cq,,~j t.o the product of the forward bias current
that the p-i–n diode resistance is definable as a and carrier lifetime (I. T ). This expression
function of frequency and depends on the dio-le applles to thin i-region p-i-n diodes ~-here the
geometry and electronic properties. diffusiun length, L r=( D T )0.5] is larger
A fnrnrula
for the p-i-n diode resistance is presente(l a~ld ttlari thf! i-region thickness, and at high
compared with experiments 1 resistance \.ersus f rcquenc ies (well above 1/ T ). A more exact
frequency data for both silicon and gallium mylat ion that applies to all p-i–n diodes can be
arsenide p–i-n diodes delived using the basic semiconductor ec~uations
irl the i-region [2] and mal be written as
Introduction
R, = kT/qI. W/L tanh (w/2L) , (~)
The forward bias resistance of a p–i-n diode
is a Primary electrical parameter that detelmitl-,s d Iere k is Boltzmann’s constant, T is the
the performance of switches and attefiuators. temyrat{ n-e and q is the electronic charee, Eqn.
Manufsdurers of p-i-n diodes specify ancl measure 2 r(?dut’es to Eqn. 1 for
forward bias resistance at a specific test W/2T, < 1. Roth equations apply only t~ the
frequency which is generally between 100 PIRz and i-l+?giorl resistance are and
high freqaency
1 GHz. Often the frequency of the application is eq~lations, meaning that all ,iunction effects (a
substantially different, both higher in the lot. fr~~quency phenomenm) are ne<lected, AS the
microwave range and lower in the RF region, from freqllency that, these de~iees are used is lowered,
this test. frequency. fiowP\ er, these ,junction effects (which can be
l?p to now, no analysis has been prese~lted 111{
xi< ~1t?d as resistances, R,I) can contribute
that characterizes the frequency dependence of s)utti fi{,ant lJ to the total p-in diode
the p-i-n diode forward resistance; that is the m si stal, ce, R? , such that RI may become only
purpose of this paper. It will be demonstrated s Smtl 1 1 part of RV in the low frequent> limit.
that the p-i-n diode resistance becomes virtuall$- mle t.(,t,al. p-i-n diode resistance as a function of
constant above a specific threshold frequ~ncy. frequency should then be vn-itten as
This frequency is definable and depends orl the
geometry of the diode, specifically the i-region it, = RI + 2R,(f). (3)
width, and its electronic properties, primariJy
the carrier lifetime and effective mobility. The TIIe factor of 2 in Eqn. 3 appears because of ths
anal ysis applies to gallium arsenide and sil.icco Lw( 1 .iurwtions (p-i and i-n) in the p-i-n diode
p-i-n diodes. (Fig. 1). Combini@ Eqns. 2 and 3, the total
The material presented derives the resistarlc~~ p-i-n di(~de resistance may ~ mitten as
of the p–i-n diode from its dc value (the slope
of the current-voltage characteristic] to its R, = kT/qI. W/L tanh (K/2L)+
value at microwave frequencies. It is applicabl(~
to packaged, beam led, and chip devi,’es. 2 R,(f), (4)
Ex~rimental data is presented of fur-ward
resistance of both gallium arsenide and si] LCOII or, in ternw of a normalized resistance ratio, 2s

p–i-n diodes of various geometries that verify


the snslysis. The paper will also indicat(c ho~. a T?, / R, ❑ 1 + 2RJ(f)/RI. (5)
100 MHz resistance value may be used to predict
the resistance at higher (or lower) frequencies,. .4n expression for R1 may be deri.:ed by
applyinq the basic semiconductor equations to the
,junct. io~ls and the i-region of the p-i–n diode.
Analysis An anal ys is relating junction and i-region
vol(.age drops in the presence of both RF and dc
The traditional high frequency i-region (h’J VF currents has been presented byReiss [3].
resistance expression, RI , is given by The result.i.ng expressions for the voltage as a
[1] flu L.-:{- i.<w] of ourrent. are quite complex but can be

RI = W/@Q , (1)

591
0149 -645 X/87/OOOO-0591$01.00 0 1987 IEEE 1987 IEEE MTT-S Digest
simplified for low level RF drive signals. with Measurements
this simplification, a frequency dependent
junction resistarrce, RJ(f), may be exqmessed as The ,measurement of the small resistance
i~alues of p-i–n dicdes and other microwave
Fb(f) = kT/qIo B tanh (W/2L) semiconductor devices was historically performed
usir~ the slotted line at a test frequency of 500
Cos (Q - (3/2 ), (6) MHz ( the practical low frequency limit for
slotted lines). Resistance was derived from high
where the variables are defined in the Appendix. SWR readi~s. The network analyzer was not
The junction resistance, RJ , arises from lrarticulary useful for . low resistance
the response of the injected carrier densit)- t<> measurements because of its ~or atacc~my
an sc sii?nal. When a low frequency ac signal is reflection coefficients close to unity;
ap~>lied to the p–i-n diode, a small ac diffusiori resistanc<?s of 0.5 Ohm to 1,0 Ohm result in
current flows through the junctions, produced b>- LW ~le( ! t ion coefficients of 0.9802 and 0.9608,
the ac \.ariationa in the dc injected carrier r<?spactively, which cannot be differentiated well
density. This response is modeled as RJ . Thr= irj a
50 Ohm network analyzer.
diffusion process is relatively slow, ho~,-ever, Olore recently, impedance analyzers, such as
and a rapidly changing ac signal will cause the KP-4191A, have become available and are
negligible variations in the dc carrier det~sit.~-, capable of performing accurate p-i-n diode
making RI the dominant resistance at high resistance measurements to values below 0.5 Ohms
frequencies. RJ and RI are affected b ~.l]e Lr] I.lw frequency range of 1.0 to 1000 mz.

i-region thickness W, carrier lifetime T, and dc ikinufacturers are now frequently using 100 MHz as
bias current 1.. This snalysis neglects tl~e the test frequency as a compromise between the
effects of package and substrate parasitic, availability of precision test fixtures and a
con tact and substrate resistance and the skin frequency compatible with the high frequency
effect, It applies only to low RF drj~-e ..e~-els [esi.stance of p-i–n diodes.
where the RF drive current is less thar! the de Measurements higher than 1.0 MHz in this
b:ias current. pal~r were performed with the RP–4191A used with
Fi.<ure 2 shows a plot of Eqn. 5 ~-ersils tk,~: a coaxial test fixture. The p-i-n diodes were
frequenc~--li fetime product for p-i-n dicdes. lradfaged in ceramic diode ~kqges, Precision
This graph applies to silicon and gall :lm} high frequency resistors were used as calibration
arsenide p–i–n diodes of any geometr>-. Th+ stsrdards and good accuracy to 1.0 GHz was
RI /R1 ratio is large at low frequencies for obtained; below 1.0 MHz down to 10 KHZ, the
small h7/L small (a W/L ratio will yield a small Hr–42i5A IA~ meter was used for resistance
Rr), thereby making the junction r(?si.stance tile measurements.
dominant 1 Ow frequency resistance. At low The carrier lifetime of p–i-n di~es is
frequencies, the total resistance approach:, t!,e traditionally measured as the reverse recovery
diode dynamic resistance [2} , which is the sl{>pe? t llne from 10 MA forward current to 6 mA initial
of the T-V characteristic, and is gi~-en b~ revvrse current. This procedure is useful as a
rel at ii-e measure of switching speed. For tkne
R,=kT/qI. [2 + W/L tanh (W/2L) 1 ❑ )m-tisluwnents presented in this paper, carrier
I I l-et i me is derived from diode stored charge [Q =
I,, T ) at fixed forward bias currents. The
Jnst rumen t,s used to measure stored charge (Bermar
which is approximately 52 ohms plus the [/[ )r}) . mod[,l 63 arid 83 stored charge meters)
i-region resistance at room temperature at ! nLi esserjt ial ly
measure the area of the rex-erse
of de bias current. At high frequencies, all th, r-t L(vi..?ry <.urrent trace ( Q .Jidt ). Stored
w/L cur~,es cent-erge t.o unit~~ (R.J goes to ::er[,) , chat gc, as the basis for carrier lifetime is also
indicating the total device resistance approarhe> bt>l t.(.r ap~>l ic.able, to p-i–n diode resistance
thf? traditional i-region resistance, R, , :$.11 arlal)-sis t ~,an reverse recovery time since it
the curves in Fig. 1 show a significant ~-ariat LOU {Ilr-vt ly relates to x-region conductivity
i n resistance in the decade near the unlt~- .al:~e moollia~ ;.m.
OffT. Eqn, G was studied for several silicon (Si)
The resistance at any frequency, Zi.{E-n t’w :UM %sllIwn arsenide (GsAs) p–i-n diodes. The
l.esistance at one frequency, may be c.>mputecl major difference in modeling these devices w-as in
using the follok-ing expression: tfle value used for the effective mobility, p .
T1l modeling Si diodes, a value Of 0.1 ~2/v.s
R~(f,))=RT(fM) [R’r/R,(fD)]/ I<as lJsed ; in the GSAS dindes, a value of o.?.o
1:,21.-. s L:as used. Figure 3 shows COrILpULed RT

[RT/RI {fM)], (8} ( l’ronl lZqn. ~) plotted versus frequency with


t~xqx:l,jmentsl data on two Si p–i-n diodes of
where fD is the desired frequency snd f~ is {Ilf ferent physical and electrical
the measured frequency of the p-i-r, ciicnie eharacterist.ics. For the thinner i-region,
resistance. This relation, with experimental shorter lifetime diode, a transition region in
data, will be discussed in the next section. tilt? 1 to 10 MHz range occurs where the dominance
of R, changes from junction resistance to

592
i-region resistance. For the thicker l-rerzlcm, (3) I?eiss, w, , “’?imdlnear Distortion .inalysls of
ionger lifetime diode, this transitlcm ree~on 1s the Ecmward Biased p-i-n Diode”, IEEE Trans.
in the 10 to 100 KHz region, Eiectrurl Devices, vol. ED-28(12), p, L49S,
l)er. 1%1.
Figure 4 S,hows ‘both experiment al 81d
analytical resistance valuee. of a GaAa p-i-n Appendix
diode versus frequency. Good agreement betwem
the analytical expression and experimental data The variables in the formula for the
is evident from the illustrated results. w7!] \ \,al(?nt junction resistance of the p-i-n diode
The total p-i-n diode resistance ZLT al]} as a f[mctltm of frequency presented in Ean. 6
frequency may be computed usin~ Eqn. 8 jr ~IN,
carrier lifetime, i-region thickness and tile,

total diode resistance at one f requenc~;-


[R,(fm] are known. For example, consid~r a
Si p–i_* diode [W=O.4 roils, Q=94pCl Ktlc+st
measured resistance value at 1 mA, 100 $I!lz 1s 8,-I
ohms . Using Eqns, 5 and 8, the prc?dlcte.j
resistance versus frequency was plotted In FIC. 5
alo~ with experimental measurements,
cm t.% s:w,,e
device (note that the analytical
curve misses
through the 100 MHz measured res~szancc la:~w .
Gocd agreement between the experimental imd
analytical results can -he seen, Irldlcati.ng the
vdldity of using Eqn. 8 in resistance
prediction.

Conclusion

The basic findings of this study demonstrate


that the total resistance of p-i-n alludes
decreases with increasing frequency. It WE

shown that p–i-n dicdes with thicker i-]eglo:ls


have less change m resistance from lo,; to high
frequency t,han thinner i-region diodes of’ th,-
same carrier lifetime,
Regarding the appl icabi 1 i.ty C)f j 00 /1}1::
resistance measurements to other frequencies, the I

followi~ guidelines should be followed:


1. If the p–i-n dicde is specified or used
in an application above 100 MHz wlere the
maximum value of resistance WLIL Ilmlt
performance (such as a switch or phase
shifter) , the 100 MHz total resistance I-alue,
RT , will be equal or larger than the higher
frequency RT value. Thus , the LOU N&
value may be considered conservat.i~e. P-or
applications below 100 MHz, the conversf
holds .

2. In attenuator applications ~,here a 1~-]-n


diode is required to be within a speclf’~ed
FIGURE 1. The p–i-n diode, showing both the
resistance r-e at a fixed curr~mt, the LOO
i-region (of length W) and the
MHz measurement may not be ‘useful at othe~-
junction regions.
frequencies. Also , thicker i-region diodes
have less variation of total resistance with
frequency than thinner ones.

References

(1) W%ite, J.F., Microwave ~emi(-K>IX+!lCtL>r

Engineering, Van Nostrand, New York, 1975.

(2) Sze, S.M., Physics of Semiconductor Devices,


John Wiley and Sons, New York, 1981.

593
0

[F;~:L
. ..._”._
-2.0
to (0-1’0 fr
10
1.0
10
a.o -0.1 ILO 10

F(MHZ)
L 0

k
I 00
0
0

I )0

FIGURE 2. Graph of the normalized resistance ratio FIGURE 4. Graph of total resistance at lmA of a
RT/R1 versus frequency-lifetime product galiium arsenide p-i-n diode versus
fT (Equation 5) for various W/L ratios frequency [W=O.12 roils, Q=8.4pC].
(applies to silicon and gallium arsenide
p-i-n diodes).

.... 0 DIODE I
— o DIODE 2
. . . .. .. .....

0 00
0
0

a
a

o

Q.. 0.
m 0000

‘%-.-%.-...*.
--w--a-”lp-.... ”...
,
lo 10 100 !000 I )00
0
F(MHZ) F(Mi-iZ)

FIGURE 3. Graph of total diode resistance at mA FIGURE 5. Graph of total resistance at lmA versus
of two silicon p-i-n diodes versus frequency using Equation 8 based on
frequency [Diode 1: W=O.42 roils, Q=l 5pc ; values of i-region length W, carrier
Diode 2: W=4.4 roils, Q=4000 PC]. lifetime T and measured resistance at
100 MHz.

594

You might also like