A page from a Russian diplomatic cable in cipher at the time of the Alaska Purchase (Wikipedia) is reproduced in Ralph E. Weber (2013), "Masked Dispatches: Cryptograms and Cryptology in American History, 1775-1900" (Internet Archive), p.119. Weber gives the source as "RG 59 Telegrams Sent by the Department of State, 1867-69, Entry 309, National Archives" (RG 59, E209, N.A.). It was a cable of 25 March 1867 (Gregorian calendar) of the negotiations from Eduard de Stoeckl, Russian minister in the United States (Wikipedia), to Alexander Gorchakov (Wikipedia), Russian Foreign Minister. It is preserved in the American archives because it was sent through the State Department telegraph office at a cost of $9,886.50 (45,000 francs). (It is believed that this does not mean the message was encrypted by an American diplomatic code, in view of the diplomatic codes/ciphers used by the US State Department at the time (see another article).)
This was during the last stage of the negotiations and the telegram encrypted the basic conditions for the purchase of Alaska for $7,000,000. A few days later, on 30 March, the purchase treaty was singed at a slightly increased price. (The text of the treaty is found at American Memory.)
The image below is the first page, covering about one third of the whole, juxtaposed with a provisional transcription (different from Weber's), in which identification of pairs 9/g, 0/o, 8/& may need correction and x, v, v indicate some markings in the image.
The plaintext of this cable is found at NARA, RG59, T1249 "Records of the Department of State Relating to Internal Affairs of the Soviet Union, 1930-1939", Roll 0029 (fold3, p.249; the table of content on p.127) among the 45 related documents found by the American embassy in Moscow in the 1930s.
The enciphering scheme is yet to be identified.
The relatively even frequency of the symbols indicate polyalphabetic substitution is employed. (The rarity of "&", "0", and "9" may suggest these should be merged with "8", "o", and "g", making the total number of different symbols match 26.)
Frequent digrams are: 2k, dg, gf, kk, p5 (4 times), followed by 4r, 8t, dk, et, j8, rg, s7, sp, w3, x2, yd, ys, yt (3 times). The only trigrams occurring twice is 7dg (at 144th and 248th places), the interval of which (104=13*2*2*2) suggests a possibility of Vigenere of keylength 13 but it does not seem to match.
A reader who has found out the scheme is kindly asked to contact the present author.
(Section added in February 2023)
I was taught by Mikhail that Russian diplomatic ciphers in the 1860s are described in a book in Russian: Tatiana Soboleva, The History of Encryption in Russia. He kindly provided me with an explanation of the systme in English. The following is what I understand from his explanation.
The system is a "biclave cipher" (which may be called a double-key cipher in English, the term sometimes also used to refer to the Vigenere cipher), invented by baron Gustav Egorovich Driesen (Густав Егорович Дризен). It is a polyalphabetic cipher based on a 26x26 enciphering table with mixed alphabets and enciphering strips to control key switching.
The 26 rows of the enciphering table are marked with 23 plaintext letters of the French alphabet (excluding k, w, and y, which are replaced with cc, vv, and ii, respectively, for enciphering) and 3 punctuation marks (dash, comma, and period). The columns are marked with 26 letters of the French alphabet, which are indexed by the key. That is, each column represents a cipher alphabet specified by the key letter. Each cipher alphabet consists of 17 letters and numbers 1, 2, 3, 4, 5, 6, 7, 8, 9 in a mixed sequence. An intersection of the row headed by the plaintext letter and the column headed by the key letter gives a cipher symbol.
Key switching is controlled by 24 key strips. Each strip has 20 key letters, and 8 of the 24 strips were used each day. That is, the sequence of 8 strips was the daily key and it was dependent on the date of encipherment. This means the key for each day was 8*20=160 letters long. When the message was longer than this, the strips (key letters) were repeated.
When enciphering, the message was processed in lines of 20 letters, divided into groups of 4 letters for convenience and aligned with 20 letters of the key strips similarly divided into groups of 4. (When a plaintext line was less than 20 letters long, it was padded.)
This polyalphabetic cipher with long (and presumably random) keys and mixed alphabets was probably among the best in the 1860s.
The Alaska purchase telegram, with letters as well as figures in the ciphertext, seems to fit this description. Although my transcription has 29 different symbols, some correction may reduce this to 26 as noted above. Then, the sample page above, consisting of 485 cipher symbols, covers three-time key repetition, with the key length assumed to be 160. Thus, if letters repeating with an interval of 160 or 320 can be identified in the plaintext, the key may be reconstructed. However, I have not been able to align the ciphertext with the plaintext, possibly because of transcription errors, uncertainty of the beginning of the ciphertext (I think it begins with "Negotiation" but am not 100% sure) or difference in handling some symbols (a hyphen, which may not occur in the ciphertext; "1825", which may be spelled out in the ciphertext).
A reader who has found out the alignment and/or the key is kindly asked to contact the present author.