Introduction to Communication Systems 1st

Edition Madhow Solutions Manual download

https://testbankdeal.com/product/introduction-to-communication-
systems-1st-edition-madhow-solutions-manual/

Explore and download more test bank or solution manual

at testbankdeal.com
We have selected some products that you may be interested in
Click the link to download now or visit testbankdeal.com
for more options!.

Contemporary Communication Systems 1st Edition Mesiya

Solutions Manual

https://testbankdeal.com/product/contemporary-communication-
systems-1st-edition-mesiya-solutions-manual/

Fundamentals of Communication Systems 1st Edition Proakis

Solutions Manual

https://testbankdeal.com/product/fundamentals-of-communication-
systems-1st-edition-proakis-solutions-manual/

Introduction to Digital Communication 2nd Edition Ziemer

Solutions Manual

https://testbankdeal.com/product/introduction-to-digital-
communication-2nd-edition-ziemer-solutions-manual/

Introduction to Law 5th Edition Hames Test Bank

https://testbankdeal.com/product/introduction-to-law-5th-edition-
hames-test-bank/
Organizational Behavior A Critical Thinking Approach 1st
Edition Neck Test Bank

https://testbankdeal.com/product/organizational-behavior-a-critical-
thinking-approach-1st-edition-neck-test-bank/

Geosystems An Introduction to Physical Geography Updated

Canadian 4th Edition Christopherson Test Bank

https://testbankdeal.com/product/geosystems-an-introduction-to-
physical-geography-updated-canadian-4th-edition-christopherson-test-
bank/

Experience Human Development 13th Edition Papalia

Solutions Manual

https://testbankdeal.com/product/experience-human-development-13th-
edition-papalia-solutions-manual/

Contemporary Issues in Accounting 2nd Edition Rankin

Solutions Manual

https://testbankdeal.com/product/contemporary-issues-in-
accounting-2nd-edition-rankin-solutions-manual/

Project Management in Practice 6th Edition Meredith

Solutions Manual

https://testbankdeal.com/product/project-management-in-practice-6th-
edition-meredith-solutions-manual/
Dosage Calculations 9th Edition Pickar Test Bank

https://testbankdeal.com/product/dosage-calculations-9th-edition-
pickar-test-bank/
 Solutions to Chapter 6 Problems
Introduction to Communication Systems, by Upamanyu Madhow

Problem 6.1
(a) Let Z = (y ∗ h)(t0 ), where h(t) = s(−t). Then Z ∼ N(m, v 2 ) if 1 sent, and Z ∼ N(0, v 2 ) if 0
sent, where Z 1
2 2 2 2 4
v = σ ||h|| = σ 4 t2 dt = σ 2
0 3
Z 1
1
Z
m = (s ∗ h)(t0 ) = s(t)s(t − t0 )dt = t(1 − t)dt =
0 6
Thus, for the ML decision rule
  r !
|m| 1 Eb
Pe = Q =Q
2v 8 N0

using the fact that we must have r

|m| Eb
=a
2v N0
(why?) where a is a constant determined by substituting Eb = 21 ||s||2 = 23 and N0 = 2σ 2 .
(b) We can improve the error probability by sampling at t0 = 0. We then have m = ||s||2 = 34 ,
while v 2 is as before. This gives, reasoning as in (a),
  r !
|m| Eb
Pe = Q =Q
2v N0

the usual formula for the performance of optimal reception of on-off keying in AWGN.
(c) For h(t) = I[0,2] , we again have the same model for the decision statistic Z = y ∗ h(t0 ), but
with v 2 = σ 2 ||h||2 = 2σ 2 , and m = (s ∗ h)(t0 ). The performance improves with |m|, which is
maximized at t0 = 2 (m = 1) or t0 = 4 (m = −1). We therefore get that, for the ML decision
rule, !
  r
|m| 3Eb
Pe = Q =Q
2v 8N0
(d) Note that we can approximate the matched filter s(−t) using linear combinations of two
shifted versions of h(t) = I[0,2] , by approximating triangles by rectangles. That is, the matched
filter shape is approximated as h̃(t) = h(t + 2) − h(t + 4). Thus, we can use the decision statistic

Z = (y ∗ h̃)(0) = (y ∗ h)(2) − (y ∗ h)(4)

We now have Z ∼ N(m, v 2 ) if 1 sent, and Z ∼ N(0, v 2 ) if 0 sent, where

v 2 = σ 2 ||h̃||2 = 4σ 2

m = (s ∗ h̃)(0) = 2
Eb
As before, we can enforce the scaling with N0
to get
  r !
|m| 3Eb
Pe = Q =Q
2v 4N0
3 dB better than the performance in (c), and 10 log10 43 = 1.25 dB worse than the optimal receiver
in (b).
Problem 6.2 (a) We have
2
e−(y−2) /18
p(y|1) = √
18π
2
e−(y+2) /8
p(y|0) = √
8π
The optimal rule consists of comparing the log likelihood ratio to a threshold. The log likelihood
ratio can be written as
   
2 1 2 1
log L(y) = log p(y|1)−log p(y|0) = −(y − 2) /18 − log(18π) − −(y + 2) /8 − − log(8π)
2 2

which has the desired quadratic form. Simplifying, we get

5y 2 + 52y + 20
log L(y) = − log 3/2
72
(b) For π0 = 1/4, we compare log L(y) to the threshold log ππ10 = − log 3. Simplifying, we obtain
the MPE rule
H1
>
5y 2 + 52y − 72 log 2 − 20 = −69.9
<
H0
(c) The conditional error probability is given by

Pe|0 = P [5Y 2 + 52Y + 69.9 > 0|H0 ]

The roots of the quadratic 5y 2 + 52y + 69.9 are α1 = −1.6 and α2 = −8.8, hence we can write
the conditional error probability as

Pe|0 = P[(Y − α1 )(Y

 −α2 ) > 0|H0] = P [Y > −1.6|H0 ] + P [Y < −8.8|H0 ]
−1.6−(−2)
=Q √
4
+ Φ −8.8−(−2)
√
4
= Q(0.2) + Φ(−3.4)
= Q(0.2) + Q(3.4)

Problem 6.3: The conditional densities and decision regions are sketched in Figure 1 (not to
scale). The threshold γ satisfies p(γ|1) = p(γ|0), or
1 2 1
√ e−γ /2 =
2π 4
p
which yields γ = 3 log(2) − log(π) ≈ 0.97.
Problem 6.4 (a) We have p(z|0) = 12 e−|z| and p(z|1) = 12 e−|z−4| , so that the log likelihood ratio,
sketched in Figure 2, is given by

p(z|1)  4, z>4
K(z) = log = |z| − |z − 4| = 2z − 4, −4 ≤ z ≤ 0
p(z|0)
−4, z<0


2
 p(z|1)

1/4 p(z|0)

−2 −γ γ 2 z

Γ1 Γ0 Γ1 Γ0 Γ1

Figure 1: Conditional densities and decision regions for Problem 6.3.

K(z)
4

0 4 z

−4

Figure 2: Log likelihood ratio for Problem 6.4.

3
(b) The conditional error probability given 1 is
R1 R1
Pe|1 = P [Z < 1|H1] = −∞ p(z|1)dz = −∞ 12 e−|z−4| dz
R1
= 21 −∞ ez−4 dz = e−3 /2 = 0.025

(c) The region z < 1 can be written as K(z) < −2. The MPE rule compares K(z) to log ππ10 .
Thus, the rule in (b) is the MPE rule if log ππ01 = −2, which yields π0 = e21+1 ≈ 0.12.

1
P(e|0)
0.9 P(e|1)

0.8

Conditional error probabilities

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
P[H0]

Figure 3: Conditional error probabilities for the MAP rule (as a function of π0 ) for Problem 6.5.

Problem 6.5 The log likelihood ratio is given by

µ1 e−µ1 y µ1
log L(y) = −µ y
= (µ0 − µ1 )y + log
µ0 e 0 µ0

where µ1 = 1/4, µ0 = 1. Plugging in, we obtain log L(y) = 43 y − log 4. The MAP/MPE rule

H1
> π0 π0
log L(y) log = log
< π1 1 − π0
H0

therefore simplifies to
H1
> 4 4π0
y log = γM AP MAPrule
< 3 1 − π0
H0
Since y ≥ 0, the MAP rule always says H1 if γM AP ≤ 0, which happens if the argument of the
4π0
log is less than (or equal to) one: 1−π 0
≤ 1, or π0 ≤ 51 .
(b) From (a), we see that for π0 ≤ 51 , the conditional error probabilities are given by Pe|0 = 1
and Pe|1 = 0, since the MAP rule always says H1 .
For π0 > 15 , we have γM AP > 0, and the conditional error probabilities are given by
 4/3
−µ0 γM AP − 34 log
4π0 1 − π0
Pe|0 = P [Y > γM AP |H0 ] = e =e 1−π0
=
4π0

4
  1/3
−µ1 γM AP − 13 4π0 1 − π0
Pe|1 = P [Y ≤ γM AP |H1 ] = 1 − e =1−e log =1−
1 − π0 4π0
These are plotted as a function of π0 in Figure 3.

1/3 p(y|1)

1/4 p(y|0)

y
Γ1 Γ0 −3/4 Γ1 3/4 Γ0 Γ1

Figure 4: ML decision regions for Problem 6.7.

Problem 6.7 (a) Densities must integrate to one. For p(y|1), he area under the triangle equals
1
2
× c × 6 = 1, so that c = 31 .
H1
>
(b) The ML rule p(y|1) p(y|0 is illustrated in Figure 4, with threshold given by equating
<
H0
conditional densities: 3 (1 − |y|/3) = 1/4, which gives |y| = 34 .
1

(c) Conditional error probabilities are given by

3 3 1 3
Pe|0 = P [|Y | < |H0] = 2 × × =
4 4 4 8
2
Pe|1 = P [|Y | > 43 |H0 ] = 2 3 13 (1 − y/3)dy
R
  4
2 2
= 23 y − y6 3 ≈ 0.45
4

The high conditional error probabilities indicate that the conditional densities are not too differ-
ent from each other, which we do see visually.

1/8 p(y|1)
1/10
p(y|0)

y
Γ1 −5 Γ0 Γ1 Γ0 5 Γ1

−γ γ

Figure 5: ML decision regions for Problem 6.8.

H1
>
Problem 6.8 (a) The ML rule p(y|1) p(y|0) is illustrated in Figure 5, with thresholds ±γ
<
H0

5
given by equating conditional densities: 18 e−|y|/4 = 1/10, which gives |y| = γ = 4 log 45 = 0.8926.
(b) Conditional error probabilities are given by
1 4 5
Pe|0 = P [|Y | < γ|H0 ] = 2 × γ × = log = 0.1785
10 5 4
R 5 1 −y/4
Pe|1 = P [γ < |Y | < 5|H0] = 2 γ 8 e dy
−γ/4 −5/4 4 −5/4
=e −e = 5 −e = 0.5135
Thus, even the ML rule can have highly asymmetric conditional error probabilities.
H1
>
noindent Problem 6.9 (a) The thresholds for the MPE rule π1 p(y|1) π0 p(y|0) with π0 are
<
H0
1 −|y|/4 5
given by (2/3) 8 e = (1/3)1/10, which gives |y| = γM P E = 4 log 2 = 3.665.
(c) Conditional error probabilities as a function of the threshold are given by the same expressions
as in Problem 6.8, but with a different value plugged in for the threshold:
1 4 5
Pe|0 = P [|Y | < γM P E |H0 ] = 2 × γM P E × = log = 0.733
10 5 2
R5
Pe|1 = P [γM P E < |Y | < 5|H0] = 2 γM P E 18 e−y/4 dy
= e−γM P E /4 − e−5/4 = 52 − e−5/4 = 0.1135
The average error probability is given by
1 2 1
Pe (MP E) = π0 Pe|0 + π1 Pe|1 = × 0.733 + × 0.1135 = 0.32 for π0 =
3 3 3
For comparison, the average error probability for the ML rule is given by averaging the conditional
error probabilities in Problem 6.8(b):
1 2 1
Pe (ML) = π0 Pe|0 + π1 Pe|1 = × 0.1785 + × 0.5135 = 0.40 for π0 =
3 3 3
Since H0 is less likely, the MPE rule lets Pe|0 get large in order to drive down Pe|1 , and achieves,
as expected, a smaller average error probability. Of course, since the hypotheses are not so well
separated to start with, the error probability for both the MPE and ML rules are rather poor.
Problem 6.10 (a) The likelihood ratio is given by
 y
p(y|1) e−m1 my1 /y! m1
L(y) = = −m0 y = e−(m1 −m0 )
p(y|0) e m0 /y! m0
The ML rule compares the log likelihood ratio to zero. Taking the log above, we have log L(y) =
m1
y log m0
− (m1 − m0 ), so that the ML rule reduces to comparing y to a threshold γ = mlog1 −m
m1 .
0
m0
For m1 = 100, m0 = 10, we have γ ≈ 39.1, so that

1, y > 39
δM L (y) =
0, y ≤ 39 = t
where t = ⌊γ⌋ in general.
(b) The conditional error probabilities are given by
t
X my 1 −m1
Pe|1 = P [Y ≤ t|H1 ] = e
y=0
y!

6
 t
X my 0 −m0
Pe|0 = 1 − P [Y ≤ t|H0 ] = 1 − e
y=0
y!

Problem 6.11 (a) The conditional densities are translations of the noise density by the signal
amplitude: p(y|0) = pN (y − A) = λ2 e−λ|y−A| and p(y|1) = pN (y + A) = λ2 e−λ|y+A| . Let us compute
and simplify the log likelihood ratio:

 2Aλ , y < −A
log L(y) = −λ|y − A| + −λ|y + A| = −2yλ , −A ≤ y ≤ A
−2Aλ , y > A


H1 H0
> >
The ML decision rule is given by log L(y) 0, and simplifies to y 0.
< <
H0 H1
(b) The conditional error probabilities are given by
Z 0 Z 0
λ −λ|y−A| λ −λ(A−y) 1
Pe|0 = P [Y < 0|H0] = e dy = e dy = e−λA
−∞ 2 −∞ 2 2
By symmetry, Pe|1 is also given by the same expression. Note that, instead of computing the
integral, we could just note that the conditional error probabilities are given by tail of an expo-
nential density (scaled down by 21 because it is two-sided).
H1
>
(c) The MPE rule is expressed in terms of the log likelihood ratio as log L(y) log ππ10 = − log 2
<
H0
1
for π0 = 3 . This corresponds to a threshold rule in y, setting −2Ay = − log 2,
H0
> 1 1
y log 2 MPE rule for π0 =
< 2λ 3
H1
(d) The LLR is the sum of the information from the observation and from the prior:
P [0|Y = y] π0 p(y|0) π0
LLR = log = log = − log L(y) + log = 2A(A/2) − log 2 = A2 − log 2
P [1|Y = y] π1 p(y|1 π1
p(y|0)
(Note that log p(y|1) = − log L(y) for our definition of likelihood ratio.)
Problem 6.12 (a) The conditional densities and ML regions are sketched in Figure 6. We
have p(y|0) = 51 e−y/5 I[0,∞) (y) and p(y|1) = 10 1
I[0,10] (y). The ML decision regions are given
by Γ1 = {y : p(y|1) > p(y|0)} and Γ0 = {y : p(y|1) < p(y|0)}, with crossover point given by
1 −y/5 1
5
e = 10 , which gives y = 5 log 2 ≈ 3.47. We can also write the ML rule as

1, 5 log 2 < y < 10
δM L (y) =
0, else
(b) The conditional error probability is given by
1
Pe|0 = P [Y ∈ Γ1 |H0 ] = P [5 log 2 < Y < 10|H0 ] = e−5 log 2/5 − e−10/2 = − e−2 = 0.3647
2

7
 1/5 p(y|0)
p(y|1)
1/10

y
Γ Γ1 Γ
0 0
5 log 2 = 3.47

Figure 6: Conditional densities and ML decision regions for Problem 6.12.

(c) The posterior probability for Y = 4 is given by

1 1 −4/5
π0 p(y|0) e
P [H0 |Y = y] = π0 p(y|0)+(1−π0 )p(y|1)
= 1
3 5
1 −4/5 2 1
e + 3 10
3 5
= 0.31

1/10 p(y|1)
Height = 2/25

y
−4 4 6 16

Base = 8

Area = conditional error probability

Figure 7: Graphical computation of conditional error probability Pe|1 for Problem 6.13(a).

 
1 |x|
Problem 6.13 We have p(y|0) = pN (y) = 1− 10
I[−10,10] and p(y|1) = pN (y − 6) =
10
 
1
10
1 − |x−6|
10
I[−4,16] . The conditional error probability Pe|1 = P [Y < 4|H1 ] is given by the area
1 2
of the triangle shown in Figure 7. The height H = p( 4|1) = 10
(1 − 2/10) = 25
, so
1 1 2 8
Pe|1 = × base × height = × 8 × =
2 2 25 25
(b) It is easy to check that the likelihood ratio is monotone nondecreasing in y. Thus, an MPE
rule will be a threshold rule in y, and the threshold is y = 4 if π0 p(y|0) = π1 p(y|1) for y = 4 for
valid prior probabilities (i.e., 0 ≤ π0 , π1 ≤ 1 with π0 + π1 = 1) Plugging in, we have
1 1
π0 (1 − 4/10) = π1 (1 − 2/10)
10 10
which simplifies
6π0 = 8π1 = 8(1 − π0 )

8
yielding a valid prior π0 = 47 .
Problem 6.14 (a) The ML rule

H1
>
hy, s1i − ||s1 ||2 /2 hy, s0i − ||s0 ||2 /2
<
H0

specializes to
H1
>
hy, s1i 0
<
H0
for s0 = −s1 . Thus, it can be implemented by correlating the received signal against s1 (t) and
then deciding
 based  on the sign of the output. The conditional error probability is given by
||s1 −s0 || R1
Pe|0 = Q 2σ
2
. Now, ||s1 − s0 || = ||2s1 || = 2||s1||, ||s1 || = 2 0 (1 − t)2 dt = 2/3, and
σ 2 = 0.1, so that !
p
2 2/3 p 
Pe|0 = Q √ =Q 20/3 = 0.0049
2 0.1
(b) Conditioned on s0 sent, y(t) = Rs0 (t)+n(t). Let us computeRthe signal and noise contributions
−0.5 −0.5
separately. The noise output n0 = −1 n(t)dt has variance σ 2 −1 12 dt = σ 2 /2. The other noise
components are independent and have identical variance, since they are computed over disjoint
intervals of equal length. Thus, ni are i.i.d., N(0, σ 2 /2). The signal contributions are given by
2
integrating s0 (t) over these intervals. It is easy to see, therefore, that y ∼ N(s0 , σ2 I), where
s0 = −(1/8, 3/8, 3/8, 1/8)T .
(c) Since the signals under the two hypotheses are negatives of each other, we infer from (b)
2
that the signal model after discretization is y ∼ N(s1 , σ2 I) conditioned on s1 sent, and y ∼
2
N(s0 = −s1 , σ2 I), where s1 = (1/8, 3/8, 3/8, 1/8)T . Thus, we are hypothesis testing in discrete
time WGN with variance σ̃ 2 = σ 2 /2. Thus, the ML rule is given by

H1
>
hy, s1 i − ||s1 ||2 /2 hy, s0i − ||s0 ||2 /2
<
H0

which specializes to
H1
1 >
hy, s1 i = (y0 + 3y1 + 3y2 + y3 ) 0
8 <
H0
Simplifying, our ML rule is
H1
>
y0 + 3y1 + 3y2 + y3 0
<
H0

9
  
The conditional error probability is given by the standard expression for WGN: Pe|0 = Q ||s12σ̃ −s0 ||
,
where we must use the variance of the discrete time WGN obtained after discretization. As be-
1 5
fore, ||s1 − s0 || = ||2s1 || = 2||s1 ||. We have ||s1 ||2 = 64 (1 + 9 + 9 + 1) = 16 , so that
p !
2 5/16
Pe|0 = Q p = Q(2.5) = 0.0062
2 0.1/2

(only a small degradation relative to ML performance for the original continuous time system).
(d) For observation Z = y0 + y1 + y2 + y3 , adding the signal components and the noise variances,
we obtain that Z ∼ N(1, 4σ̃ 2 = 2σ 2 ) if s1 is sent, and Z ∼ N(−1, 2σ 2 ) if s0 is sent. This falls
within the basic Gaussian format, and the conditional error probability is given by
√
   
|m1 − m0 | 1 − (−1)
Pe|0 = Q =Q √ = Q( 5) = 0.0127
2v 2 2σ 2
(e) Smallest error probability is in (a), largest is in (d). We could have rank ordered without
explicit computation. The probability in (a) is the best achievable, since we are implementing the
ML rule for the original continuous time observation. We lose some performance in discretizing as
in (b), because we are not doing optimal processing of the continuous time signal (which we know
is to correlate with s1 ). We lose further performance in (d), because we are not implementing
the ML rule for the discretized signal y.
Problem 6.15 Drawing figures is highly recommended, but we leave that as an exercise.
(a) The filter output at time t is given by
Z Z t
z(t) = y(τ )h(t − τ )dτ = y(τ )dτ
t−1

Thus, the filter is basically doing integrate-and-dump, and adding filter samples is equivalent to
a correlation of y against an appropriately chosen waveform. Thus,
Z 0 Z 1 Z 1
Z = z(0) + z(1) = y(τ )dτ + y(τ )dτ = y(τ )dτ = hy, ui
−1 0 −1

where the “effective correlator” u(t) = I[−1,1] (t). Conditioned on 0 sent, y(t) = n(t) and Z =
hn, ui ∼ N(0, σ 2 ||u||2 = 2σ 2 ). Conditioned on 1 sent, y(t) = s(t) + n(t) and Z = hs, ui + hn, ui =
A + hn, ui ∼ N(A, 2σ 2 ).
(b) The model in (a) falls within the standard Gaussian format, and the ML rule is given by

H1
> A
Z
< 2
H0

The error probability is  

A−0
Pe = Pe|0 = Pe|1 = Q √
2 2σ 2
A2
Noting that Eb = 12 ||s||2 = 3
, we obtain
√  r !
3Eb 3Eb
Pe = Q √ =Q
2 N0 4N0

10
Visit https://testbankdead.com
now to explore a rich
collection of testbank,
solution manual and enjoy
exciting offers!
This is a factor of 3/4 (or about 1.25 dB) worse than optimal demodulation of OOK.
(c) The decision statistic
Z 0 Z 0.5 Z 1
Z2 = z(0) + z(0.5) + z(1) = y(τ )dτ + y(τ )dτ + y(τ )dτ = hy, vi
−1 −0.5 0

where the effective correlator v(t) = I[−1.−0.5] (t)+2I[−0.5,0.5](t)+I[0.5,1] (t) (draw this to see how the
shape compares with the original signal s(t)). We leave it as an exercise to show that ||v||2 = 5
and hs, vi = 47 A. Following the same steps as in (b), we have Z2 ∼ N(0, σ 2 ||v||2 = 5σ 2 ) if 0 is
sent, and Z2 ∼ N(hs, vi = 74 A, σ 2 ||v||2 = 5σ 2 ) if 1 is sent. This again falls within the standard
Gaussian format, and the error probability is given by
7 7
√ !
A − 0 3E

b
Pe = Pe = Pe|0 = Pe|1 = Q 4 √ =Q p4
2 5σ 2 2 5N0 /2
which simplifies to !
r
147Eb
Pe = Q
160N0
This is a factor of 147/160 (or 0.37 dB) worse than OOK with optimal demodulation. Thus,
the degradation is smaller than in (b), which is to be expected since the shape of the effective
correlator v(t) here is a better approximation for the signal s(t) than the effective correlator u(t)
in (b).

a(t)

t
1 2 3
Figure 8: Correlator for implementing ML rule in Problem 6.16(a).

Problem 6.16 (a) The ML rule

H1
>
hy, s1i − ||s1 ||2 /2 hy, s0i − ||s0 ||2 /2
<
H0
can be implemented by a single correlator as follows:
H1
> ||s1||2 − ||s0 ||2
hy, s1 − s0 i
< 2
H0

11
The correlator a(t) = s1 (t) − s0 (t) is sketched in Figure 8. We have ||s1 ||2 = 2 and ||s0 ||2 =
R1 2 2
2 0 t2 dt = 2/3, so that the threshold γM L = ||s1 || −||s
2
0 ||
= (2 − 2/3)/2 = 2/3.
1 2 2 1
(b) Note that Eb = 2 (||s1 || + ||s0 || ) = 2 (2 + 2/3) = 4/3 and

d2 = ||s1 − s0 ||2 = ||s1||2 + ||s0 ||2 − 2hs1 , s0 i = 2 + 2/3 − 2(−1/2) = 5/3

d2
so the power efficiency Eb
= 45 . The ML rule has error probability
s ! r !
d2 Eb 5Eb
Pe = Q =Q
Eb 2N0 6N0

(c) The MPE rule is given by

H1
>
hy, s1i − ||s1||2 /2 + σ 2 log π1 hy, s0i − ||s0 ||2 /2 + σ 2 log π0
<
H0
and can be implemented using the same correlator as the ML rule, but with the threshold
adjusted to account for the priors:
H1
> ||s1 ||2 − ||s0 ||2 π0 π0
hy, s1 − s0 i γM P E = + σ 2 log = γM L + σ 2 log
< 2 π1 π1
H0

For π0 = 13 , the threshold is therefore given by γM P E = γM L − σ 2 log 2 = 2/3 − 0.693σ 2.

R(d) For h(t) = I[0,1] (t), filter-and-sample is the same as integrate and dump: (y ∗ h)(t) =
t−1
ty(τ )dτ . The samples at t = 1, 2, 3 correspond to different intervals of unit length, hence the
noise outputs are i.i.d. N(0, σ 2 ). The signal outputs (si ∗ h)(t) = t−1 ts(τ )dτ for t = 1, 2, 3 are
R

easily computed and form the vectors: s1 = (0, 1, 1)T and s0 = (−1/2, −1/2, 0)T . The conditional
distribution of Z given 0 sent is therefore N(s0 , σ 2 I).
(e) Actually, since the noise is uncorrelated, this is not really challenging as promised. This is
simply signaling in discrete WGN, and the ML rule is given by
H1
>
hZ, s1 i − ||s1 ||2 /2 hZ, s0 i − ||s0 ||2/2
<
H0
which simplifies to
H1
>
hZ, s1 − s0 i h(||s1 ||2 − ||s0 ||2 )/2
<
H0
Plugging in the numbers, we obtain
H1
1 3 >
Z1 + Z2 + Z3 3/4
2 2 <
H0

12
The error probability for ML reception is discrete WGN is given by
s !
d2 Eb
 
||s1 − s0 ||
Pe = Q =Q
2σ Eb 2N0

where d2 = ||s1 − s0 ||2 = 7/2 is for the discretized system, but Eb = 4/3 for the original
continuous time system, computed in (a). (Note that the scaling is such that the noise variance
per dimension is the same in both systems, otherwise we would have to account for that.) We
have d2 /Eb = 21/8, so that !
r
21Eb
Pe = Q
16N0

Problem 6.17
(a) Signal space representations with respect to the given orthonormal basis are:
Signal Set A: s1 = (1, 0, 0, 0)T , s2 = (0, 1, 0, 0)T , s3 = (0, 0, 0, 1)T and s4 = (0, 0, 0, 1)T
Signal Set B: s1 = (1, 0, 0, 1)T , s2 = (0, 1, 1, 0)T , s3 = (1, 0, 1, 0)T and s4 = (0, 1, 0, 1)T
(b) For Signal Set A, the pairwise distance between any two points satisfies d2 = d2min = 2, while
the energy per symbol is Es = 1. Thus, Eb = Es /(log2 4) = 1/2, and d2min /Eb = 4. The union
bound on symbol error probability is therefore given by
q  p 
p
2
Pe (signal set A) ≤ 3Q dmin /Eb Eb /2N0 = 3Q 2Eb /N0

For signal set B, each signal has one neighbor at distance given by d21 = 4 and two at distance
given by d22 = d2min = 2. The energy per symbol is Es = 2, so that Eb = 1. The union bound is
given by
p p  p p 
Pe (signal set B) ≤ 2Q d22 /Eb Eb /2N0 + Q d21 /Eb Eb /2N0
p  p 
= 2Q Eb /N0 + Q 2Eb /N0

(c) For exact analysis of error probability for Signal Set B, suppose that the received signal in
signal space is given by Y = (Y1 , Y2 , Y3 , Y4 ). Condition on the first signal s1 = (1, 0, 0, 1)T being
sent. Then
Y1 = 1 + N1 , Y2 = N2 , Y3 = N3 , Y4 = 1 + N4
where N1 , ..., N4 are i.i.d. N(0, σ 2 ) random variables. A correct decision is made if hY, s1i >
hY, sk i, k = 2, 3, 4. These inequalities can be written out as

hY, s1i = Y1 + Y4 > hY, s2i = Y2 + Y3

hY, s1i = Y1 + Y4 > hY, s3i = Y1 + Y3
hY, s1i = Y1 + Y4 > hY, s4i = Y2 + Y4

The second and third inequalities give Y1 > Y2 and Y4 > Y3 , and imply the first inequality. Thus,
the conditional probability of correct reception is given by
Pc|1 = P [Y1 > Y2 and Y4 > Y3 |1] = P [Y1 > Y2 |1]P [Y4 > Y3 |1]
since Yk are conditionally independent given the transmitted signal. Noting that Y1 − Y2 and
Y4 − Y3 are independent N(1, 2σ 2 ), we have
  2    
1 1 2 1
Pe|1 = 1 − P c|1 = 1 − 1 − Q √ = 2Q √ −Q √
2σ 2 2σ 2 2σ 2

13
 1
p
Setting √2σ 2
= a Eb /N0 , with Eb = 1 and N0 = 2σ 2 , we have a = 1. Further, by symmetry, we
have Pe = Pe|1 . We therefore obtain that
p  p 
2
Pe = 2Q Eb /N0 − Q Eb /N0 , exact error probability for signal set B

1) 3)

t t
−2 −1 0 1 2 −2 −1 0 1 2

2) 4)

t t
−2 −1 0 1 2 −2 −1 0 1 2

Figure 9: Natural basis for signal set in Problem 6.18, labeled in the order of the coordinates
they correspond to in signal space.

Problem 6.18 Using the natural basis shown in Figure 9, the 4 signals map (not worrying
about scale factors to the vectors a = (1, 1, 1, 1)T , b = (1, 1, −1, −1)T , c = (1, −1, 1, −1)T ,
d = (1, −1, −1, 1)T . It is easy to check that these vectors are all orthogonal, so this is 4-ary
equal energy, orthogonal signaling. Thus, Es ≡ 4 (hence Eb = logEs 4 = 2) and all pairwise
2
distances are given by d2 = 2Es = 8. The power efficiency d2 /Eb = 4.
(a) The union bound is given by
dcb
Pe|c ≤ Q(
q
dca
2σ
) + Q( Q( d2σ
 2σ ) +q
cd d
)= 3Q( 2σ )
2
d Eb 2Eb
=Q Eb 2N0
=Q N0

(b) False: not more power-efficient than QPSK. The power efficiency equals 4, which is the same
as that of QPSK.
Problem 6.19
(a) For 8-PSK, the symbol energy Es = R2 . For the QAM constellations,

Es (QAM1) = I − channel avg energy + Q − channel avg energy

= 84 [(d1 /2)2 + (3d1 /2)2 ] + (d1 /2)2 = 32 d21

Es (QAM2) = I − channel avg energy + Q − channel avg energy

= [ 68 (d2 /2)2 + 28 (3d2 /2)2 ] + [ 48 d22 + 84 0] = 54 d22
(1) (2)
where d1 = dmin and d2 = dmin . Since the number of constellation points is the same for each
constellation, equal energy per bit corresponds to equal symbol energy, and occurs when
p p
R = 3/2 d1 , d2 = 6/5 d1

(b) From (a), for the same Eb ,

π
d8P
min
SK
= 2R sin = .937d1 < d1 < d2
8

14
so that, in the high SNR regime, we expect that

Pe (8P SK) > Pe (QAM1) > Pe (QAM2)

(c) Each symbol is assigned 3 bits. Since 8-PSK and QAM1 are regular constellations with at
most 3 nearest neighbors per point, we expect to be able to Gray code. However, QAM2 has
some points with 4 nearest neighbors, so we definitely cannot Gray code it. We can, however,
try to minimize the number of bit changes between neighbors. Figure 10 shows Gray codes for
8-PSK and QAM1. The labeling for QAM2 is arbitrarily chosen to be such that points with 3
or fewer nearest neighbors are Gray coded.
011 010
101
000 010 110 100
111 001
011 000 001 110 100
R 000
001 011 111 101
010 100 101 111
110
8-PSK QAM1 QAM2

Figure 10: Bit mappings for Problem 6.19.

(d) For Gray coded 8-PSK and QAM1, a symbol error due to decoding to a nearest neighbor
causes only 1 out of the 3 bits to be in error. Hence, using the nearest neighbors approximation,
P [bit error] ≈ 13 P [symbol error]. On the other hand, P [symbol error] ≈ N̄dmin Q( dmin
2σ
), where
N̄dmin is the average number of nearest neighbors. While the latter is actually an upper bound on
the symbol error probability (the nearest neighbors approximation coincides with the intelligent
union bound in these cases), the corresponding expression for the bit error probability need not
be an upper bound (why?).
For 8-PSK, dmin = 2R sin π8 and Es = 3Eb = R2 . Plugging in σ 2 = N0 /2 and N̄dmin = 2, we
obtain s 
1
2 3(1 − 2 )Eb
√
P [bit error]8P SK ≈ Q  
3 N0

For QAM1, N̄dmin = 5/2 and Es = 3Eb = 3d21 /2, so that

r !
5 Eb
P [bit error]QAM 1 ≈ Q
6 N0

For QAM2, we need to make nearest neighbors approximation specifically for bit error probability.
Let N(b) total number of bit changes due to decoding into nearest neighbors when symbol b is
1
P
sent. For the labeling given, these are specified by Table 1. Let N̄bit = 8 b N(b) = 11/4 denote

b 000 001 010 011 100 101 110 111

N(b) 1 6 2 2 1 2 6 2
Table 1: Number of bit changes due to decoding into nearest neighbors for each symbol of QAM2

the average number of bits wrong due to decoding into nearest neighbors. Since each signal
point is labeled by 3 bits, the nearest neighbors approximation for the bit error probability is
now given by !
  r
1 dmin 11 6Eb
P [bit error]QAM 2 ≈ N̄bit Q = Q
3 2σ 12 5N0

15
Other documents randomly have
different content
very large surface and by the free circulation of air. It is possible to
make wine vinegar by the quick process, but it is not done, because
the product is inferior in taste and aroma to that made by the slow
process.

Both wine vinegar and malt vinegar when freshly prepared have a
stupefying and unpleasant odour. Before the product is ready for the
market, it has to be matured in barrels. During this process, a small
quantity of alcohol which still remains in the vinegar combines slowly
with some of the acetic acid, producing acetic ester, a substance
which has a pleasant fruity odour.

The colour of wine vinegar is natural, but vinegar which is produced

by the quick process is colourless or only faintly coloured. Since the
public has a preference for vinegar which is brown in colour, the
product of the quick process is coloured artificially, either by adding
caramel or by preparing the weak spirit from malt which has been
slightly charred in drying.

Industrial Acetic Acid. The solutions of acetic acid dealt with

above would be too dilute for any industrial purpose; moreover, the
acid can be obtained much more cheaply by the distillation of wood.
When wood is subjected to a high temperature, it is converted into
charcoal and, at the same time, an inflammable gas, an acid liquid,
and tar are given off, and can be collected in suitable vessels. The
following table, on page 73, gives the relative amounts of the
various substances obtained from wood by dry distillation. The
quantities are those derived from one cord, that is, 125 cu. ft.

73

Charcoal Alcohol Calcium Tar in Wood oil Turpentine

in in acetate gallons. in gallons.
bushels. gallons. in lbs. gallons.
Hard 40-50 8-12 150-200 8-20
woods
Resinous 25-40 2-4 50-100 30-60 30-60 Heavy
woods woods 12-
25
Light
woods 2-
10
Sawdust 25-35 2-4 45-75

The aqueous liquid that distils over contains methyl alcohol (wood
spirit), acetone, and acetic acid. The crude mixture is known as
pyroligneous acid. This is neutralized with milk of lime or soda ash,
which converts acetic acid into calcium or sodium acetate, but has
no action on the methyl alcohol and acetone which are also present.
The mixture is then distilled, when methyl alcohol, acetone, and
water pass over into the distillate, leaving the acetate in the retort.

To obtain the free acid from the acetate, the latter is well dried and
then distilled with concentrated sulphuric acid. Acetic acid, being the
more volatile of the two acids, distils over, and is nearly pure.

The method of removing the last traces of water depends upon the
fact that acetic acid solidifies at 17° C. The acid, which is nearly, but
not quite, free from water, is cooled until a portion solidifies. The
part which still remains liquid is poured away, and the process is
repeated until a residue is obtained which solidifies as a whole. This
is glacial acetic acid, so called because it is a mass of glistening
plates which look like newly-formed ice.

The Acetates

Aluminium Acetate, made by dissolving alumina in acetic acid, is

the “red liquor” which is used as a mordant in dyeing. It is a
colourless liquid, but is called “red liquor” because it is used with
dyes which give a red colour.
Ferrous Acetate, made in a similar way from scrap iron and 74
acetic acid, is the “black liquor” used in dyeing.

Verdigris, or basic copper acetate, is a valuable pigment. It is made

by interposing cloths soaked in vinegar between plates of copper.
After the action has been allowed to go on for a long time, the
plates are washed with water and the verdigris is scraped off. The
finest verdigris is made in France in the wine-producing district
around Montpellier. Here, instead of cloths soaked in vinegar, the
solid residue from the wine presses is spread in layers between the
copper plates. The product made in this way is called vert de
Montpellier.

Fig. 11. DUTCH PROCESS FOR WHITE LEAD

Verdigris, like all the copper compounds, is extremely poisonous. It

is very liable to be formed on the surface of copper utensils used for
cooking purposes.
Lead Acetate, or sugar of lead, is used in large quantities in the
colour industry for making various reds and yellows. It is 75
prepared by dissolving the metal or its oxide (litharge) in
acetic acid.

The slow action which acetic acid vapour has upon the metal lead
finds a very interesting application in what is known as the Dutch
[4]
process for the manufacture of white lead for paint. The metal is
cast into grids or spirals, which are placed on the shoulders of the
specially made pots sketched in Fig. 11. A little dilute acetic acid is
poured into each of the pots, which are then arranged side by side
on a thick layer of tan bark, stable manure, or other material which
will heat by fermentation. The first layer of pots is then boarded
over; another layer of pots is placed upon this, and so on, tier upon
tier, until the shed is quite full. The heat developed by the
fermenting material vaporizes the acetic acid, and this vapour
corrodes the lead, forming basic lead acetate. The carbon dioxide
which is also produced during fermentation converts the acetate into
the carbonate, which falls as a heavy white powder into the pots.

Future Supply of Acetic Acid. When all the operations involved in

the production of acetic acid from wood, from the felling of the tree
to the final separation of the glacial substance, are taken into
consideration, it will be readily understood how it is that this acid
has never been cheap when compared with other acids used on an
equally large scale. In addition to this, the competition for wood for
paper-making and for the very numerous cellulose industries is
rapidly increasing. It is, therefore, not surprising to learn that
chemists have turned their attention towards the discovery of newer
and cheaper methods of making acetic acid.

Such a process seems to have been worked out in Germany. 76

The starting-point is acetylene gas made by the action of
water on calcium carbide. When this gas is passed through sulphuric
acid containing suspended mercuric oxide or dissolved mercury salt,
the acetylene is oxidized first to aldehyde and then to acetic acid.

If this process should prove to be successful, it will form the

starting-point of a new and important industry, for, apart from the
large amount of acetic acid which is used in commerce, there is the
production of the very important solvent known as acetone, which
can be made from acetic acid by a very simple operation.

Tartaric Acid. Grape juice contains a large quantity of potassium

hydrogen tartrate dissolved in it; when the liquid is fermented and
alcohol is formed, this salt crystallizes out because it is not soluble in
alcohol. After the new wine has been poured off, the salt is found as
a brownish crystalline residue adhering to the sides of the vat. Also
the salt goes on crystallizing after the wine is put into barrels, and
forms an incrustation on the sides. This is called the lees or
sediment of wine. In commerce, the substance is known as argol
(sometimes spelt argal), and also tartar of wine.

Crude argol is purified by dissolving it in water and destroying the

colour by boiling with animal charcoal. When the clear liquid
obtained from this mixture by filtration is evaporated, a white
crystalline substance separates out. This is potassium hydrogen
tartrate or cream of tartar.

Tartaric acid is obtained from cream of tartar. The salt is dissolved in

water and nearly neutralized with milk of lime. Insoluble calcium
tartrate is precipitated, and potassium tartrate remains in solution. A
further quantity of calcium tartrate is obtained by adding 77
calcium chloride to the solution just mentioned. The two
precipitates of calcium tartrate are then mixed and decomposed by
dilute sulphuric acid, and after the calcium sulphate is filtered off,
tartaric acid is obtained as a solid by evaporating the clear liquid.

The general properties of tartaric acid are well known. It is soluble in

water, giving a solution which has a pleasantly acid taste.
Citric Acid. The sharp flavour of many unripe fruits is due to the
presence of citric acid; the juice of lemons contains 5-6 per cent. of
the acid. The free acid is obtained in a manner precisely similar in
principle to that described for tartaric acid.

Oxalic Acid. Oxalic acid and its salts, the oxalates, are very widely
distributed in the vegetable kingdom. These compounds are present
in wood sorrel (Oxalis acetosella), in rhubarb, in dock, and in many
other plants. The acid is made on a large scale by mixing pine
sawdust to a stiff paste with a solution containing caustic soda and
potash. The paste is spread out on iron plates and heated, care
being taken not to heat the mixture to the point at which it chars.
The mass is then allowed to cool, and is mixed with a small quantity
of water to dissolve out the excess of alkali. This is recovered and
used again.

Sodium oxalate, which is the main product of the reaction described

above, is dissolved in water and treated with milk of lime, whereby
insoluble calcium oxalate is obtained, which is subsequently
decomposed with sulphuric acid, yielding oxalic acid.

Potassium hydrogen oxalate is sometimes called salts of sorrel, and

potassium quadroxalate, salts of lemon. The most familiar use of the
latter substance is in the removal of ink stains.

Oxalic acid and its salts are poisonous. The free acid has 78
sometimes been mistaken for sugar with fatal results.

Formic Acid (L. formica, an ant) is found both in the vegetable and
in the animal kingdom. If the leaf of a stinging nettle is examined
with a microscope, it is seen to be covered with long pointed hairs
having a gland at the base. This gland contains formic acid. When
the nettle is touched lightly, the fine point of the hair punctures the
skin, and a subcutaneous injection of formic acid is made, which
quickly raises a blister.
The inconvenience which arises from the stings of bees and wasps,
also from the fluid ejected by ants when irritated, is due to formic
acid. The remedy in each case is the same; the acid must be
neutralized as quickly as possible with mild alkali, such as washing
soda.

Formic acid was first made by distilling an infusion of red ants. It is

now made from glycerine and oxalic acid.

The Fatty Acids. Animal fats and vegetable oils are similarly
constituted bodies. They are composed mainly of three chemical
compounds known as stearin, palmitin, and oleïn. Of these, stearin
and palmitin are solids at ordinary temperatures, while oleïn is a
liquid. Hard fats like those of mutton and beef are composed mainly
of stearin; fats of medium hardness contain stearin, palmitin, and
some oleïn; while oils such as cod-liver oil and olive oil are nearly
pure oleïn.

Stearin, palmitin, and oleïn are analogous in composition to salts.

Their proximate constituents are glycerine and certain organic acids,
stearic, palmitic, and oleïc respectively.

In order to obtain the fat free from tissue which it contains in its
natural state, it is tied up in a muslin bag and heated in boiling
water. The fat is squeezed out through the meshes of the fabric and
floats on the surface of the water as an oil which solidifies on 79
cooling. This clarified fat is called tallow.

All fats and vegetable oils can be resolved into their two
constituents, the acid and the glycerine. This can be brought about
by heating the fat with water to about 200° C. This operation must
be carried out in a vessel capable of withstanding pressure and
closed with a safety valve; otherwise, the requisite temperature
could not be obtained. After this treatment, there is left in the vessel
an oily layer which solidifies on cooling and an aqueous layer which
contains the glycerine. The solidified oily layer is the fatty acid. In
the case of mutton or beef tallow, it would be mainly a mixture of
stearic and palmitic acids. This mixture is used to make “stearin”
candles. The acids themselves are wax-like solids without any
distinctive taste. Stearic acid melts at 69° C. and palmitic at 62° C.
They have no perceptible action on the colour of litmus, neither have
they any solvent action on metals or carbonates. We should not
recognize these substances as acids at all were it not for the fact
that they combine with alkalis, forming salts.

The salts of the fatty acids are called soaps. To make soap, the fat is
boiled with caustic alkali or caustic lye, as it is more often called.
This breaks the fat up primarily into the acid and glycerine; but in
this case, instead of obtaining the acid as the final product as we did
above by heating with water under pressure, we get the sodium or
potassium salt of the acid according to the alkali used. When caustic
soda is used, the product is a hard soap; when caustic potash is
used, it is a soft soap. The treatment of fats in this way with caustic
alkalis is called “saponification.”
80
 CHAPTER VIII
MILD ALKALI
Caustic and Mild. There are two classes of alkalis distinguished by
the terms caustic and mild. If a piece of all-wool material is boiled
with a solution of caustic soda or potash, it dissolves completely,
giving a yellow solution. Mild alkali will not dissolve flannel, though it
may have some slight chemical action causing shrinkage. Partly for
this reason, and partly because commercial washing soda often
contains a little caustic soda, woollen garments must not be boiled
or even washed in hot soda water.

The disintegrating action of the caustic alkalis is also illustrated by

the use of caustic soda in the preparation of wood pulp for paper
making. Tree trunks are first torn up and shredded by machinery;
but notwithstanding the power of modern machinery, the fibre is not
nearly fine enough for the purpose until it has been “beaten” with a
solution of caustic soda, whereby the pulp is brought to a smooth
and uniform consistency like that of thin cream.

Mild Soda and Potash. Until the middle of the eighteenth century,
it was thought that the soluble matter extracted from the ashes of
all plants was the same. In 1752 it was shown that the substance
obtained in this way from plants which grew in or near the sea
differed from that from land vegetation by producing a golden yellow
colour when introduced into the non-luminous flame of a spirit lamp,
while that from land plants gave to the flame a pale lilac tinge. The
former substance is now known as mild soda, and the latter as mild
potash.
At this point it is well to make it clear to the reader that there 81
are two bodies commonly called soda, and two called potash.
One of each pair is caustic and one mild.

By a simple chemical test it is easy to distinguish a mild from a

caustic alkali. When a little dilute acid is added to the former, there is
a vigorous effervescence caused by the escape of carbon dioxide,
but no gas is given off when a caustic alkali is treated in the same
way. The liberation of carbon dioxide on the addition of acids shows
that the mild alkalis are carbonates.

Washing Soda is so well known, that very little description of its

external characteristics is necessary. It is a crystalline substance,
easily soluble in water. The crystals, when freshly prepared, are
semi-transparent; but after exposure to air for some time, they are
found to lose their transparency and to become coated with an
opaque white solid which crumbles easily. This change in appearance
is accompanied by a loss in weight.

Crystals of soda melt very easily on the application of heat and, on

continued heating, the liquid seems to boil. When this operation is
carried out in a vessel attached to a condenser, the vapour that is
given off from the melted soda condenses to a clear colourless liquid
which, on examination, proves to be water. When no more water
collects in the receiver, the vessel contains a dry, white solid, which
by any chemical test that may be applied is shown to be the same as
washing soda, but it contains no water of crystallization and has a
different crystalline form. This substance is anhydrous sodium
carbonate, or soda ash as it is called in commerce. When soda ash is
mixed with water, it combines with about twice its own weight of
that liquid, forming soda crystals again.

Washing soda, then, contains nearly two-thirds of its weight of 82

water. Some of this water is given off spontaneously when the
soda is exposed to air; the water may even be said to evaporate.
This accounts for the loss of weight observed and also for the
formation of the white layer of partially dehydrated soda over the
surface of the crystal. The property of losing water in this way is
common to most crystals containing a high percentage of water of
crystallization. The phenomenon is known as “efflorescence.” It may
here be observed that crystals of washing soda which have become
coated over in this way contain relatively more soda than those
which are transparent.

Natural Soda. In Egypt, Thibet, and Utah, there are tracts of

country where the soil is so impregnated with soda that the land is
desert. The separation of the soda from the earth is a simple
operation, for it is only necessary to agitate the soil with water and,
after the insoluble matter has settled down, to evaporate the clear
solution until the soda crystallizes out.

In addition to alkali deserts, there are also alkali lakes. Those in

Egypt are small, nevertheless, about 30,000 tons of soda per annum
are exported from Alexandria. Owens Lake in California is said to
contain sufficient soda to supply the needs of North America; while
in the East African Protectorate, beneath the shallow waters of Lake
Magadi (discovered in 1910), there is a deposit of soda estimated at
200,000,000 tons.

The Leblanc Process. At the present time, the greater part of the
world’s supply of soda is made from common salt by two processes.
The older of these, which is known as the Leblanc process, was
introduced in France towards the end of the eighteenth century. In
those days soda was very dear, for the main supply came from the
ashes of seaweeds; wherefore the French Academy of Sciences, in
1775, offered a prize for the most suitable method of converting salt
into soda on a manufacturing scale. The prize was won by Nicholas
Leblanc, who in 1791 started the first soda factory near Paris. These
were the days of the French Revolution; the “Comité de Sûreté
Général” abolished monopolies and ordered citizen Leblanc to
publish the details of his process.
 83

Fig. 12. SALT CAKE FURNACE

The first alkali works were established in Great Britain in 1814. 84

The total amount of soda now made in this country every year
is about 1,000,000 tons, of which nearly one-half is still made by the
Leblanc process.

Salt Cake. The first stage of the Leblanc process consists in mixing
a charge of salt weighing some hundredweights with the requisite
amount of “chamber” sulphuric acid. The operation is carried out in
a circular cast-iron pan (D, Fig. 12) about 9 ft. in diameter and 2 ft.
deep. The pan is covered over with a dome of brickwork, leaving a
central flue (E) for the escape of hydrochloric acid gas which is
produced. At first, the reaction takes place without the application of
heat, but towards the end the mass is heated for about one hour.
The contents of the pan are then raked out on to the hearth of a
reverberatory furnace (a, b) and more strongly heated. More
hydrochloric acid gas is given off, and the reaction is completed. The
solid product which remains is impure Glauber’s salt (sodium
sulphate), and is known in the trade as “salt cake.”

Black Ash. In the second stage of the Leblanc process, salt cake is
converted into black ash. The salt cake is crushed and mixed with an
equal weight of powdered limestone or chalk and half its weight of
coal dust. This mixture is introduced into a reverberatory furnace
(Fig. 13) by the hopper K, and heated to about 1000° C. by flames
and hot gases from a fire at a. During this operation, the mass is
kept well mixed, and after some time it is transferred to h where the
temperature is higher. The mixture then becomes semi-fluid and
carbon monoxide gas is given off.

85

Fig. 13. BLACK ASH FURNACE

The formation of carbon monoxide within the semi-solid mass 86

renders it porous. This is an advantage, because it greatly
facilitates the subsequent operation of dissolving out the soluble
sodium carbonate. The appearance of the flames of carbon
monoxide at the surface of the black ash indicates the end of the
process. The product is then worked up into balls and removed from
the furnace.

The chemical changes which take place in making black ash are
probably as follows: Carbon (coal dust) removes oxygen from
sodium sulphate, which is thus changed to sodium sulphide. This
substance then reacts with the limestone (calcium carbonate),
forming sodium carbonate (soda) and calcium sulphide.

Extraction of Soda. It now only remains to dissolve out the soda

from the insoluble impurities with which it is mixed in the black ash.
It is evident that the smaller the amount of water used for this
purpose the better, because the water has subsequently to be got rid
of by evaporation. The process of extraction is, therefore, carried out
systematically. The black ash is treated with water in a series of
tanks which are fitted with perforated false bottoms. The soda
solution, which is heavier than water, tends to sink to the bottom
and, after passing through the perforations, is carried away by a
pipe to the second tank, and so on throughout the series. The fresh
water is brought first into contact with the black ash from which
nearly all the soda has been extracted.

The method of finishing off the black ash liquor differs 87

according to the final product which the manufacturer desires
to obtain, for the liquor contains caustic soda as well as mild soda.
For the present, we will suppose that the end product is to be
washing soda. In this case, carbon dioxide is passed into the liquor
to convert what caustic soda there is into mild soda.

The clarified soda liquor is then evaporated until crystals of soda

separate out. The first part of this process is carried out in large
shallow pans (P. Fig. 13), using the waste heat of the black ash
furnace, and finally in vats containing steam-heated coils. As the
crystals separate out, they are removed, drained, and dried.

Alkali Waste. Black ash contains less than half its weight of soda,
so that for every ton of soda produced there is from a ton and a half
to two tons of an insoluble residue which collects in the lixiviating
and settling tanks. This residue is known as alkali waste.

Alkali waste is of no particular value. It is not even suitable as a

dressing for the land, and since it is not soluble in water there is no
convenient means of disposing of it. Consequently, it is just
accumulated at the works and, as the heap grows at an alarming
rate, it cumbers much valuable ground. Moreover, it contains
sulphides from which, under the influence of air and moisture,
sulphuretted hydrogen is liberated. Alkali waste, therefore, has a
very unpleasant odour.

The whole of the sulphur which was contained in the sulphuric acid
used in the first stage of the process remains in the alkali waste,
mainly as calcium sulphide. A plant for the recovery of this sulphur is
established in some of the larger works. The alkali waste is mixed
with water to the consistency of a thin cream, in tall, vertical
cylinders. Carbon dioxide under pressure is forced into the mixture,
and this converts the calcium sulphide into calcium carbonate 88
and sets free hydrogen sulphide, which, when burnt with a
limited supply of air, yields sulphur.

By this process, the most unpleasant feature of alkali waste, namely,

the smell, is removed. The calcium carbonate which remains is of
very little value. Some of it is used in making up fresh charges for
the black ash process and some for preparing Portland cement, for
which finely-ground calcium carbonate is required; the remainder is
thrown on a heap.

Bicarbonate of Soda. Bicarbonate of soda can be easily

distinguished from washing soda. It is a fine, white powder similar in
appearance to the efflorescence on soda crystals. It does not contain
any water of crystallization.

When bicarbonate of soda is heated, it does not melt, and, as far as

its external appearance is concerned, it does not seem to undergo
any change. If, however, suitable arrangements are made, water and
carbon dioxide gas can be collected, and the sodium bicarbonate will
be found to have lost 36·9 per cent. of its weight. The substance
which remains is identical with that obtained by heating soda
crystals, that is, anhydrous sodium carbonate. Sodium bicarbonate
is, therefore, a compound of sodium carbonate and carbonic acid.

The most familiar use of this compound is indicated by its common

names “baking-soda” and “bread-soda.” It is mixed with dough or
other similar material in order to keep this from settling down to a
hard solid mass in baking. The way in which bicarbonate of soda
prevents this will be readily understood when it is remembered that
an ounce of this substance liberates more than 2,300 cu. in. of
carbon dioxide when it is heated. When the bicarbonate of soda is
well mixed with the ingredients of the cake or loaf and disseminated
throughout the mass, each particle will furnish (let us say) its bubble
of gas. Since these cannot escape, a honey-combed structure is
produced.

89
Fig. 14. THE SOLVAY PROCESS
Baking powder is a mixture of bicarbonate of soda and ground 90
rice; the latter substance is merely a solid diluent.

The Solvay Process. Soda ash is one of the principal forms of mild
alkali used in commerce. Large quantities of this substance are made
by heating bicarbonate of soda. We shall now consider another alkali
process in which this substance is the primary product.

For the greater part of the first century of its existence, the Leblanc
soda process had no rival, although another method, known as the
ammonia-soda process, was patented as early as 1838. In this case,
however, as in many others, expectations based on the experiments
carried out in the laboratory were not realized when the method
came to be tried under manufacturing conditions. It was not until
1872 that Ernest Solvay, a Belgian chemist, had so far solved the
difficulties, that a new start could be made. In that year, about 3,000
tons of soda were produced by the ammonia-soda or Solvay process,
as it has now come to be known. Since then, however, the quantity
produced annually has been steadily increasing, until at the present
time it amounts to more than half of the world’s supply.

The Solvay process is very simple in theory. Purified brine is

saturated first with ammonia gas and then with carbon dioxide.
Water, ammonia, and carbon dioxide combine, forming ammonium
bicarbonate, which reacts with salt (sodium chloride), producing
sodium bicarbonate and ammonium chloride.

The principal reaction is carried out in a tower (Fig. 14 (1), a, a)

from 50 to 65 ft. in height and about 6 ft. in diameter. At 91
intervals of about 3½ ft. throughout its length, the tower is
divided into sections by pairs of transverse discs, one flat with a
large central hole, and one hemispherical and perforated with small
holes (Fig. 14 (2)). The discs are kept in position by a guide rod G.
Fig. 14 (3) shows a better arrangement of the guide rods. In modern
works, the space between the discs is kept cool by pipes conveying
running water. The ammoniated brine is led into the tower near its
middle point. The carbon dioxide is forced in at E in the lowest
segment, and as it passes up the tower it is broken up into small
bubbles by the sieve plates. Sodium bicarbonate separates out as a
fine powder, which makes its way to the bottom of the tower
suspended in the liquid.

The perforated plates are necessary for the proper distribution of

carbon dioxide through the brine. They are, however, a source of
trouble, because the holes quickly become blocked up with sodium
bicarbonate, and every ten days or so it is necessary to empty the
tower and clean it out with steam or boiling water.

Recovery of Ammonia. The production of 1 ton of soda ash by

the Solvay process involves the use of a quantity of ammonia which
costs about eight times as much as the price realized by selling the
soda. It is evident that the success of the process as a commercial
venture depends largely on the completeness with which the
ammonia can be recovered.

During the process, ammonia is converted into ammonium chloride,

which remains dissolved in the residual liquor. From this ammonia
gas is set free by adding quicklime and by blowing steam through
the mixture. It is now claimed that 99 per cent. of the ammonia
used in one operation is recovered.

Soda Ash. The bicarbonate of soda produced by the Solvay 92

process is moderately pure. For all ordinary purposes, it is only
necessary to wash it with cold water to remove unchanged salt, and
after drying, it is ready to be placed on the market if it is to be sold
as bicarbonate. The greater part of the Solvay product, however, is
converted into soda ash by the application of heat. If soda crystals
are required, the soda ash is dissolved in water and crystallized.

In many ways, the Solvay process compares very favourably with the
older method. It is an advantage to start with brine, for that is the
form in which salt is very often raised from the mines. The end
product is relatively pure; moreover, it is quite free from caustic
soda, which for some purposes for which soda ash is used is a great
recommendation. There is no unpleasant smelling alkali waste. On
the other hand, the efficiency of the Solvay process is not high, for
only about one-third of the salt used is converted into soda. This
would make the process impossible from the commercial point of
view were it not for the cheapness of salt.

The Leblanc process, too, has its advantages. In the next chapter we
shall see that it is adaptable for the production of caustic as well as
mild alkali. The chlorine which is recovered in the Leblanc process is
a very valuable by-product. In the Solvay process, chlorine is lost, for
hitherto no practicable method has been found for its recovery from
calcium chloride.

The position with regard to the future supply of alkali is very

interesting. The competition between the Leblanc and the Solvay
processes for supremacy in the market is very keen. At the same
time, both processes are in some degree of danger of being
supplanted by the newer electrical methods, which will be mentioned
in the last chapter.

The following table shows very clearly the rapid progress 93

made by the Solvay process in ten years. The quantities are
given in tonnes (1 tonne = 0·9842 ton).

1884. 1894.
Leblanc Solvay Leblanc Solvay
soda. soda. soda. soda.
Great Britain 380,000 52,000 340,000 181,000
Germany 56,500 44,000 40,000 210,000
France 70,000 57,000 20,000 150,000
United States — 1,100 20,000 80,000
Austria- 39,000 1,000 20,000 75,000
Hungary
Russia — — 10,000 50,000
Belgium — 8,000 6,000 30,000
545,500 163,100 456,000 776,000

Mild Potash. Potassium carbonate (mild potash) was formerly

obtained from wood ashes. The clear aqueous extract was
evaporated to dryness in iron pots, and the substance was on this
account called potashes; later, potash. A whiter product was
obtained by calcining the residue, and this was distinguished as
pearl-ash. Chemically pure potassium carbonate was formerly
obtained by igniting cream of tartar (potassium hydrogen tartrate)
with an equal weight of nitre. It is for this reason that potassium
carbonate is sometimes called “salt of tartar.”

About the middle of last century, natural deposits of potassium

chloride were discovered in Germany. The beds of rock salt near
Stassfurt are covered over with a layer of other salts, and for many
years these were removed and cast aside as “waste salts”
(abraumsalze). When at a later date they were examined more
carefully, they were found to contain valuable potassium compounds,
notably the chloride. After that discovery, mild potash was 94
made by the Leblanc process., and Germany controlled the
world’s markets for all potassium compounds.

At the outbreak of war, the German supplies of potassium

compounds ceased as far as the allied nations were concerned, and
an older method of making potassium chloride from orthoclase or
potash-felspar was revived. This involves the heating of the
powdered mineral to a high temperature after mixing it with calcium
chloride, lime, and a little fluorspar. The potassium chloride is then
extracted from the fused mass with water. This method has been
worked with great success in America, and it is claimed that
potassium chloride can be made in that country at a cost which is
lower than that formerly paid for the imported article.
Mild potash and soda are so very similar in chemical properties that
in most cases it is immaterial which compound is used. In all cases
in which there is this choice, soda is employed, both because it is
cheaper and because it is more economical, for 106 parts of soda
ash are equivalent to 138 parts of potash. There are, however, some
occasions when soda cannot be substituted, notably for the
manufacture of hard glass and soft soap, and for the preparation of
caustic potash, potassium dichromate, and other potassium salts.

Potassium Bicarbonate. This resembles the corresponding sodium

salt in nearly every respect. It is, however, much more readily
soluble in water, so much so, that it is not possible to obtain this
substance by the Solvay method. It is made from potassium
carbonate by saturating a strong aqueous solution of that substance
with carbon dioxide.

95
 CHAPTER IX
CAUSTIC ALKALIS
The Alkali Metals. The discovery of current electricity in 1790
furnished the chemist with a very powerful agency for bringing
about the decomposition of compounds. Hydrogen and oxygen were
soon obtained by passing an electric current through acidulated
water; and in 1807, Sir Humphry Davy, who is perhaps better
remembered for his invention of the miners’ lamp, isolated the
metals sodium and potassium by subjecting caustic soda and caustic
potash respectively to the action of the current.

Sodium and potassium are very remarkable metals. They are only a
little harder than putty, and can easily be cut with a knife or
moulded between the fingers. When exposed to the air, they rust or
oxidize very rapidly, so much so that they have to be preserved in
some mineral oil or in airtight tins. They are lighter than water,
which they decompose with the liberation of hydrogen, and under
favourable circumstances the hydrogen takes fire so that the metals
appear to burn on the surface of the water. After the reaction is over
and the sodium or potassium has disappeared, a clear colourless
liquid remains which has a strongly alkaline reaction, and when this
is evaporated until the residue solidifies on cooling, caustic soda or
potash is obtained. For very special purposes, the caustic alkalis are
sometimes made by the action of the metals on water, but for
production on a large scale, less expensive methods are adopted.

Caustic Alkali is obtained from the corresponding mild alkali 96

in the following way. The substance—washing soda, for
example—is dissolved in water and the solution is warmed. Lime is
stirred into this solution, and from time to time a small test portion
of the clear supernatant liquid is removed and mixed with a dilute
mineral acid. When this ceases to cause effervescence, the change is
complete. The clear liquid is now separated from the solid matter
(excess of lime together with calcium carbonate) and evaporated in
a metal dish. Since the caustic alkalis are extremely soluble in water,
they do not crystallize as do most of the compounds previously
described. Evaporation is, therefore, carried on until the liquid which
remains solidifies when cold.

Caustic Soda. To describe the process by which caustic soda is

manufactured, we must return to the making of black ash. The
mixture from which black ash is made contains limestone. It is
heated to 1000° C., which is a sufficiently high temperature to
convert limestone into lime. When the black ash is subsequently
treated with water, the lime which is present converts some of the
mild alkali to caustic; consequently, black ash liquor always contains
both alkalis.

When the manufacturer intends to make caustic soda and not soda
crystals, the composition of the black ash mixture is varied by adding
a larger proportion of limestone, so that there may be an excess of
lime in the black ash produced. The treatment with water is carried
out as described under washing soda, and then more lime is added
to convert the mild soda into caustic soda. After the excess of lime
and other suspended matter has settled down, the clear caustic
liquor is evaporated in iron kettles until it becomes molten caustic,
which will solidify on being allowed to cool.

There are various grades of caustic soda on the market 97

differing one from another in purity. The soap manufacturer
uses caustic liquor or lye containing about 40 per cent. of caustic
soda. For other purposes, the solid containing from 60 to 78 per
cent. is used. Sometimes the product is whitened by blowing air
through the strong caustic liquor or by the addition of a little
potassium nitrate. Finally, for analytical purposes, caustic soda is
purified by dissolving it in alcohol and subsequently evaporating the
clear liquid.

Caustic Potash. The methods for the preparation of the

corresponding potassium compound are precisely the same as those
described for caustic soda; in fact, wherever the words sodium and
soda occur in this chapter, the reader can always substitute
potassium and potash respectively.

Caustic Lime. Apart from its use in making mortar and cement,
lime is very often employed to neutralize acids. For this purpose, a
suspension in water, called milk of lime, is generally used, for lime
itself is not very soluble. Probably it is only the soluble part which
reacts; nevertheless, as soon as this is used up, more of the solid
dissolves, and in this way the action goes on as if all the lime were
in solution.

Lime is also a very valuable substance in agriculture, especially on

damp, boggy land, where there is much decaying vegetable matter,
and on land which has been liberally manured. The soil in these
cases is very likely to become acid and is then unproductive. Lime is
added to “sweeten” the soil; in other words, to neutralize the acid.

Ammonia. The pungent smelling liquid popularly known as “spirits

of hartshorn” is a solution of ammonia gas in water. It is a caustic
alkali and, as such, is sometimes used to remove grease spots. Here,
however, we shall consider ammonia only in connection with 98
ammonium salts, some of which are used in very large
quantity as fertilizers.

The principal source of ammonia at the present time is the

ammoniacal liquor obtained as a by-product in the manufacture of
gas for heating and lighting. Coal contains about 1 per cent. of
nitrogen, and when it is distilled, some of this nitrogen is given off as
ammonia, which dissolves in the water produced at the same time.
This liquid is condensed in the hydraulic main and in other parts of
the plant where the gas is cooled down.

Gas liquor contains chiefly the carbonate, sulphide, sulpho-cyanide,

and chloride of ammonia, together with many other substances,
some of which are of a tarry nature. It would not be practicable to
evaporate this liquid with a view to obtaining the ammonium salts,
because it is only a very dilute solution. Hence, after the removal of
tar, the liquor is treated in such a way that ammonia is set free.

In some cases the liberation of ammonia is accomplished by blowing

superheated steam into the liquor, which sets free the ammonia
which is combined as carbonate, sulphide, and sulpho-cyanide, but
not that which is present as chloride. In other works, the gas liquor
is mixed with milk of lime, which liberates all the combined
ammonia. The ammonia is then expelled from the mixture by a
current of steam or air and steam. In both cases, the gas which is
given off is passed into sulphuric acid, whereby ammonium sulphate
is formed in solution and afterwards obtained as a solid by
evaporation.

Ammonium Salts

Ammonium Chloride. Like all other alkalis, ammonia solution

neutralizes acids, forming salts. With hydrochloric acid, it produces
the white solid known as sal ammoniac or ammonium chloride. 99
This compound is familiar as the one required to make the
liquid used in a Leclanché cell, which is generally used as the current
generator for electric bells.

Ammonium Carbonate, which is also called stone ammonia and

salt of hartshorn, is made by subliming a mixture containing two
parts chalk and one part ammonium sulphate. It is a white solid
which gives off ammonia slowly and is, therefore, used as the basis
for smelling salts.
Ammonium Nitrate is obtained by passing ammonia gas into nitric
acid until it is neutralized. It is a white solid, which melts easily on
being heated, and breaks up into water and nitrous oxide (laughing
gas), which is the “gas” administered by dentists. Ammonium nitrate
is also used in the composition of some explosives: for example,
“ammonite” is said to contain 80 per cent. of this substance.

Ammonium Sulphate is used chiefly as an artificial manure; the

amount required for this purpose throughout the world is over
1,500,000 tons every year.

Synthetic Ammonia. Though the soluble compounds of nitrogen

are fairly abundant, the supply is by no means equal to the demand,
because such enormous quantities are required for agricultural
purposes. It has been already said that ammonia is obtained as a
by-product in the distillation of coal, and it has been repeatedly
pointed out that our coal supplies are far from inexhaustible;
moreover, coal gas may not always be used for lighting and heating.
It, therefore, becomes a very important question as to how the
future supply of ammonium salts is to be maintained.

Ammonia is a very simple compound formed from the elements

nitrogen and hydrogen, and, as before mentioned, the supply of free
nitrogen in the air is literally inexhaustible. In recent years, 100
the efforts of chemists have been directed towards finding a
method of converting the free nitrogen of the air into some simple
soluble compound. This problem is usually spoken of as the “fixation
of nitrogen.”

In the Haber process, nitrogen obtained by the fractional distillation

of liquid air is mixed with three times its volume of hydrogen, and
this mixture is heated to between 500°C. and 700°C. under a
pressure of 150 atmospheres (nearly 1 ton to the square inch) and
in the presence of a contact agent. Under these conditions, nitrogen
and hydrogen combine to form ammonia, which is condensed by
passing the mixed gases into a vessel cooled with liquid air, any
unchanged nitrogen and hydrogen being passed back again over the
contact substance.

The problem of making ammonia from the air is closely connected

with that of making nitric acid from the same source. In some
experiments the two are combined, and ammonium nitrate is
produced directly. Ammonia made by the Haber process, or some
modification, is mixed with atmospheric oxygen and passed through
platinum gauze heated to low redness. This results in the formation
of nitric oxide, which is further oxidized by atmospheric oxygen; and
finally, from a mixture of oxides of nitrogen, water vapour, and
ammonia, synthetic ammonium nitrate is obtained.

101
 CHAPTER X
ELECTROLYTIC METHODS
One of the most noteworthy developments of modern chemical
industry has been the increasing use of electricity as an agent for
bringing about changes in matter. This has followed naturally from
the reduction in the cost of electricity, due in great measure to the
utilization of natural sources of energy which for untold ages had
been allowed to run to waste.

This last achievement is likely to produce such a change in economic

conditions that it is worth while giving a little thought to what may
be called a newly-discovered asset of civilization. One example will
make this clear. In the bed of the Niagara river, which flows from
Lake Erie to Lake Ontario, there is a sudden drop of 167 ft. over
which the water rushes with tremendous force and expends its
energy in producing heat which cannot be utilized. This is a waste of
energy, but it cannot be circumvented because no method has yet
been found to control the waters of the Falls themselves.
Nevertheless, by leading the head waters through suitable channels
from the high level to the low, it is possible to use the energy to
drive turbines, which, in their turn, drive dynamos which produce
the current. This is merely the conversion of the energy of running
water into electrical energy; and while the sun remains, this supply
of energy will be forthcoming in undiminished quantity, because by
the heat of the sun the water is lifted again as vapour, which
descends as rain to replenish the sources from which the Niagara
flows.

Electricity is employed in chemical industry in two ways. In 102

the first place, it may be used to produce very high
temperatures required for the reduction of some metallic ores, for
melting highly-refractory substances, and for making steel. It is,
however, rather with the second method, called electrolysis, that we
are here mainly concerned.

Fig. 15. THE ELECTROLYSIS OF SALT SOLUTION

Solutions of acids, bases, and salts, and in some cases the fused
substances themselves, conduct the electric current; but at the same
time they suffer decomposition. This method of decomposing a
substance is known as electrolysis, or a breaking up by the agency
of electricity.
The apparatus required in a very simple case is shown in Fig. 15. It
merely consists of some suitable vessel to contain the liquid; two
plates—one to lead the current into the solution, the other to lead it
away again—and wires to connect the plates to the poles of a
battery, storage-cell, or dynamo. Each plate is called an 103
electrode, and distinguished as positive or negative according
as it is joined to the positive or negative pole of the current
generator. By convention, electricity is supposed to “flow” from the
positive pole of the battery to the positive electrode or anode, and
then through the solution to the negative electrode or cathode, and
so back to the negative pole of the generator, thus completing the
circuit external to the battery.

When acids, alkalis, and salts are dissolved in water, there is strong
evidence to show that they break up to a greater or less extent into
at least two parts called ions. These are atoms, or groups of atoms,
[5]
which have either acquired or lost one or more electrons. They
move about quite independently of one another and in any direction
until the electrodes are placed in the liquid. Then they are
constrained to move in two opposing streams—those which have
acquired electrons all move towards the negative electrode, and
those which have lost electrons towards the other. At the electrodes
themselves, the former give up and the latter take up electrons, and
become atoms again. Let us now consider a concrete example.
Common salt is composed of atoms of sodium and atoms of chlorine
paired. When a small quantity of this substance is dissolved in a
large quantity of water, the pairing no longer obtains. The chlorine
atoms move away independently accompanied by an extra satellite
or electron, and the sodium atoms move away also but with their
electron strength one below par. When the current is introduced into
the liquid, the sodium ions travel towards the cathode and chlorine
ions towards the anode, and when they reach the goal, sodium ions
gain one electron and chlorine ions lose one, and both become
atoms again. Chlorine atoms combine in pairs forming molecules and
escape from the solution in the greenish yellow cloud that we call
chlorine gas. The sodium atoms react immediately with water,
forming caustic soda with the liberation of hydrogen.

To return now to practical considerations. The electrolysis of 104

salt solution appears to be an ideally simple method of
obtaining caustic soda and chlorine from sodium chloride. As a
manufacturing process, it would seem to be perfect, for the salt is
broken up directly into its elements and a secondary reaction gives
caustic soda automatically. There is no “waste” as in the Leblanc
process, and it does not require the use of any expensive
intermediary substance afterwards to be recovered, as in the Solvay
process. But, as very often happens when working on a large scale,
difficulties arise, and these up to the present have only been
partially overcome.

Some of the chlorine remains dissolved in the liquid and reacts with
the caustic soda, forming other substances which, though valuable,
are not easy to separate from the caustic soda. It is possible to get
over this difficulty to some extent by placing a porous partition
between the anode and the cathode, and in that way dividing the
cell into cathodic and anodic compartments. As long as the partition
is porous to liquids, it will allow the current to pass, but at the same
time it will greatly retard the mixing of the contents of the two
compartments. Porous partitions or cells which are in common use
for batteries are made of “biscuit” or unglazed porcelain.

It must be remembered, however, that porous partitions only retard

the mixing of liquids; they do not prevent it. Moreover, a 105
further difficulty arises from the fact that chlorine is a most
active substance, and therefore it is difficult to find a material which
will resist its corrosive action for any length of time, and the same
difficulty arises in the case of the anode where the chlorine is given
off.

Castner Process for Caustic Soda. The following is the most

successful electrical process for the manufacture of caustic soda yet
devised. It was introduced in 1892, and is known as the Castner
process. It should be noted that the use of the porous partition has
been avoided in a very ingenious way.

Fig. 16. THE CASTNER PROCESS

The cell (see Fig. 16) is a closed, rectangular-shaped tank divided

into three compartments by two non-porous partitions fixed at one
end to the top of the tank, while the other end is free and fits
loosely into a channel running across the tank. The floor of the tank
is covered with a layer of mercury of sufficient depth to seal the
separate compartments. The two end compartments contain the
brine in which are the carbon anodes; the middle compartment
contains water or very dilute caustic soda in which the cast-iron
cathode is immersed.

The current enters the end compartments by the carbon 106

anodes and passes through the salt solution to the mercury
layer which in these compartments are the cathodes. The current
then passes through the mercury to the middle compartment, and
then through the solution to the cathode, thence back to the
Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.

More than just a book-buying platform, we strive to be a bridge

connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.

Join us on a journey of knowledge exploration, passion nurturing, and

personal growth every day!

testbankdeal.com